Sentinel Docs

Product documentation for the Sentinel legal compliance & eDiscovery platform

Use your browser's Print → Save as PDF to export this document.

Home

Product Documentation

Master the record.

Everything you need to ingest, search, review, and act on a matter's evidence — and to use Sentinel's AI assistant with confidence. User guides, an API reference, security details, and answers, all in one place.

Get started → Core concepts API reference

Welcome to the documentation for Sentinel — an enterprise platform for legal compliance monitoring and eDiscovery. Read it top to bottom as a manual, or use the sidebar search (press /) to jump to a topic. Every page has a print link, and a single printable version bundles the whole manual into one document for PDF export.

Ingestion

Upload files, email archives, and media. Sentinel extracts text, runs OCR, deduplicates, and indexes everything automatically.

Search

Keyword, semantic, hybrid, and visual search across a matter's documents and transcripts — find by exact term or by meaning.

Emma, the AI assistant

An assistant grounded in your evidence — searches, drafts answers with verifiable citations, and helps with review, by text or voice.

Document review

A three-pane workspace with coding, tagging, bulk actions, email threading, and predictive coding (TAR/CAL).

Sessions

Record and transcribe depositions, interviews, and meetings with speaker diarization and live AI assistance.

Security & compliance

Per-tenant isolation, data residency in your Azure environment, audit logging, and citation-verified AI output.

Where to start

  • IntroductionWhat Sentinel is & who it's for
  • Getting StartedSign in and find your way around
  • Core ConceptsThe whole mental model, one page
  • AdministrationUsers, roles, settings, audit
  • API ReferenceAuth, endpoints & the AI connector
  • FAQ & GlossaryQuick answers and definitions

Practice modes

Sentinel adapts its interface to your practice. A tenant runs in one of three practice modes — Litigation, Transactional, or Real Estate — which changes the navigation and available features. See Core Concepts → Practice modes.

A note on accuracy. This documentation describes the Sentinel product as built. Some features are practice-mode-specific or role-gated, and a small number of surfaces are still in active development — these are called out where relevant. If something in your environment differs, your administrator may have configured it differently, or you may be on a different practice mode.

Introduction

Introduction

Sentinel is an enterprise SaaS platform for legal compliance monitoring and eDiscovery. It gives legal and compliance teams one place to ingest large volumes of documents and communications, make them instantly searchable, review and code them, and surface the facts that matter — supported throughout by an AI assistant that is grounded in your own evidence rather than the open web.

Who Sentinel is for

  • Litigation teams managing discovery: collecting custodian data, reviewing productions, logging privilege, building productions, and preparing for depositions, hearings, and trial (including jury selection).
  • Corporate / transactional teams running deals: organizing diligence data rooms, tracking document request lists, and working due‑diligence checklists.
  • Real‑estate teams managing property transactions with deal pipelines, diligence, and closing checklists.
  • Compliance and investigations teams monitoring communications and documents against regulatory frameworks.

What makes Sentinel different

Everything is searchable, by meaning. Every file that enters Sentinel flows through a single ingestion pipeline that extracts text (running OCR on scans and multimodal extraction on images, audio, and video), deduplicates it, and builds a searchable index. There are no feature‑specific "silos" that the search layer can't see — a document you upload to a data room, a transcript from a deposition, and an email pulled from a monitored mailbox are all findable the same way. See Core Concepts → The content pipeline.

An AI assistant you can trust. Emma, the built‑in assistant, answers questions about your matter using only what your documents actually say. When Emma references a document she does so with a citation token that is mechanically verified against the documents a tool actually returned — if a response references a document that wasn't retrieved, the system blocks it. This "citation grounding" is core to keeping AI output defensible as legal work product. See Emma → Citations and guardrails.

Your data stays in your environment. Sentinel is multi‑tenant by design, but each customer's documents live in that customer's own database and storage. AI calls that leave the environment for inference are configured to not retain your data. See Security & Compliance.

The shape of the product

At a high level, you work in Sentinel like this:

  1. Pick a matter. A matter is a case (litigation), a deal (transactional), or a property transaction (real estate). It is the container for all of that work's documents, sessions, findings, and analysis.
  2. Get documents in. Upload them, share an upload link with a client, import a court docket, or connect a monitored mailbox. Sentinel ingests and indexes everything automatically.
  3. Find and review. Search by keyword or meaning, open documents in the review workspace, tag and code them, and ask Emma questions.
  4. Produce the work. Build Bates‑stamped productions, maintain a privilege log, draft motions, work checklists, prepare for depositions, and export.

Practice modes at a glance

The interface adapts to one of three practice modes per tenant:

Mode Primary objects Representative features
Litigation Cases, custodians Discovery review, productions, privilege log, motions, jury selection, depositions
Transactional Deals, counterparties Deal data rooms, request lists, due‑diligence checklists, business intelligence
Real Estate Properties, stakeholders Deal pipeline, diligence, closing checklists

Practice mode is set per tenant by an administrator (and by Sentinel during provisioning). See Core Concepts → Practice modes.

How to use this documentation

  • The Getting Started section gets you signed in and oriented.
  • Core Concepts is the single best page to understand the whole system.
  • The User Guides each cover one feature in depth.
  • Administration, Integrations, API Reference, and Security & Compliance are reference material for admins, integrators, and evaluators.
  • The FAQ and Glossary answer common questions and define terms.

Continue to Getting Started →

Getting Started

Getting Started

This guide gets you signed in to Sentinel and oriented in the interface.

Signing in

Open your tenant's Sentinel URL (your administrator or welcome email provides it) and sign in with your email and password.

Sentinel separates the login / account service (the control plane) from the application. When you sign in, the control plane verifies your identity and issues your session; the application then loads with your tenant's data.

First sign‑in (new accounts)

If your account was just created — for example, you were invited by an administrator or received trial credentials — your first sign‑in may walk you through a short setup:

  1. Set a new password. Accounts created with a temporary password are prompted to choose a permanent one on first login. You can't reuse the temporary password.
  2. Enroll in two‑factor authentication (2FA). If your tenant requires 2FA, you'll be guided to set up an authenticator app (scan the QR code, enter the 6‑digit code). After enrollment you'll enter a code from your authenticator each time you sign in.

Trouble signing in

  • Forgot password — use the Forgot password link on the login page to receive a reset email.
  • Lost your 2FA device — ask a tenant administrator to reset your 2FA (see Administration → Users & roles).
  • "Subscription expired" / account locked — trial or lapsed accounts see an informational page directing you to your account contact. Reach out to your administrator or the address shown on that page.

The Home hub

After signing in you land on a Home hub designed to get you moving quickly. It includes:

  • An omnibox where you can type a question or command for Emma, or start a search.
  • Quick actions — start a live session, open research/chat, browse documents.
  • Briefing signals — at‑a‑glance counts for things like matters, ingestion progress, alerts, and your next calendar event.
  • Recent matters — pick up where you left off.

Selecting a matter

Most work in Sentinel happens inside a matter (a case, deal, or property transaction). Use the matter selector in the sidebar to choose your current matter. Once selected, the sidebar expands to show that matter's workspace — overview, documents/data room, review, analysis, and so on — and the URL reflects the selected matter so you can bookmark or share deep links.

You can switch matters at any time from the same selector. To see everything you have access to, open the Matters list (called Deals or Deal Pipeline in transactional / real‑estate modes).

Finding your way around

The left sidebar is your primary navigation. What you see there depends on two things:

  1. Your tenant's practice mode (Litigation, Transactional, or Real Estate) — this determines whole sections of the app. See Core Concepts → Practice modes.
  2. Your role — administrators see administration tools that other users don't. See Roles overview below.

Common top‑level destinations:

  • Home — the orientation hub.
  • Sessions — recorded depositions, interviews, and meetings.
  • Privileged Chat / Research — the unified Emma chat and research surface.
  • Matters / Deals — your list of cases or deals.
  • My Library / Firm Library — personal and firm‑wide document collections.
  • Calendar — scheduling with relationship briefing.
  • Help — in‑app walkthroughs and guidance.
  • Administration — (admins) users, settings, API keys, audit log, and more.

When a matter is selected, matter‑scoped sections appear below — for example Data Room, Doc Review, Productions, Analytics, and analysis tabs.

Roles overview

Your role controls what you can see and do. The tenant‑level roles are:

Role What it's for
Admin Tenant administrator — manages users, settings, API keys, and has access to every matter in the tenant.
Attorney Case/deal team member. Sees the matters they are assigned to (or that are org‑wide). The default role for most users.
Deal Team Transactional / M&A team member focused on deals and due diligence.
Platform Admin Sentinel staff role for platform operations across tenants — not a customer‑assigned role.

Within a matter, an attorney's specific access level (full, read‑only, or limited) is set when an admin assigns them to that matter. See Core Concepts → Access control.

Next steps

  • Read Core Concepts for the full mental model.
  • Jump into a feature: Data Rooms & Uploading Documents, Search & Research, or Emma — the AI Assistant.

Core Concepts

Core Concepts

This page is the single best place to understand how Sentinel fits together. Everything else in the documentation builds on these ideas.

Matters: the container for work

A matter is the top‑level container for a piece of work and everything related to it — documents, sessions, findings, analysis, and team access. What a matter represents depends on your practice mode:

  • Litigation → a case (with custodians, a court docket, discovery, and productions).
  • Transactional → a deal (with counterparties, a data room, and diligence).
  • Real Estate → a property transaction (with stakeholders and closing diligence).

You always work in the context of a selected matter. The matter selector in the sidebar sets your current matter; the URL reflects it so you can bookmark and share deep links.

There are also special matter types:

  • Personal data lake — a personal workspace owned by a single user.
  • Enterprise data lake — an organization‑wide collection that every user in the tenant can access.

Documents: data rooms and case files

Documents in a matter live in two complementary surfaces:

  • Data Room — the matter's primary document repository: received evidence, discovery documents, and deal documents. In Sentinel's model, the matter is its data room, so "the data room" and "the matter's documents" are effectively the same set.
  • Case Files — firm‑authored work product kept distinct from discovered evidence: your own drafts, research, pleadings, and notes.

Both are display/grouping layers over the same underlying indexed content (see below), so search and Emma see everything regardless of which surface a document came in through.

The content pipeline

The most important architectural idea in Sentinel: every file that enters the system goes through one unified ingestion pipeline and becomes searchable. There is no feature that stores content "off to the side" where search can't see it.

When a file is uploaded (or arrives via a monitored mailbox, a client‑upload link, or a docket import), it flows through a multi‑stage pipeline:

  1. Parse — the file is identified by type. Plain documents go straight to text extraction; email containers (PST, MBOX, MSG) and archives (ZIP) are sent to the split stage.
  2. Split — archives and email containers are decomposed into their individual files and attachments, each of which re‑enters the pipeline.
  3. OCR — for scanned PDFs and images with no extractable text, Sentinel runs real optical character recognition (Azure Document Intelligence) to recover the text.
  4. Multimodal extraction — for images, audio, video, and image‑bearing PDFs, a multimodal AI model extracts meaning that plain text extraction would miss (e.g. the content of a photo, a diagram, or a scanned signature).
  5. Embed — the extracted content is split into chunks and converted into vector embeddings that power meaning‑based (semantic) search.
  6. Complete — the job is finalized: counts are tallied, the search index is rebuilt, and the document is fully searchable.

Two ideas underpin this:

  • One canonical content record with deduplication. Each unique file's extracted text and metadata are stored once, keyed by a content hash (SHA‑256). If the same file appears again — say, the same memo received in two productions — Sentinel reuses the existing extraction instead of re‑processing it.
  • Embeddings for semantic search. The chunked embeddings are stored in a vector index so that a query like "emails discussing regulatory pressure" finds documents by meaning, not just by exact words.

What you see while a document is processing

Ingestion is asynchronous, so a freshly uploaded document moves through visible states such as queued, processing / extracting (parse, OCR, or multimodal extraction), embedding, and complete — at which point it appears in search results. Files that can't be processed (unsupported or empty) are marked failed/skipped with a reason you can review. Large uploads and media files naturally take longer than small text documents.

Search

Sentinel offers several search modes over a matter's content (and transcripts):

  • Keyword — exact terms, phrases, names, Bates numbers, and Boolean operators. Fast and precise.
  • Semantic — meaning‑based search using embeddings; finds conceptually relevant documents even when they don't use your exact words.
  • Hybrid — combines keyword precision with semantic recall; a good default for natural‑language questions.
  • Visual — searches the visual content of images and scanned pages (signatures, handwriting, diagrams, photos) using multimodal embeddings.

See Search & Research for how to use them, and Emma for asking questions in natural language.

Emma: the AI assistant

Emma is Sentinel's built‑in AI assistant. She can search your documents, answer questions about a matter, analyze communications, navigate the app, code and tag documents, and summarize sessions — by text or by voice.

Emma's defining trait is citation grounding: she may only reference a document using a citation token that a tool actually returned during the conversation, and her responses are mechanically checked against those tokens. If a response references a document that wasn't retrieved, it is blocked rather than shown. This keeps AI output grounded in your real evidence. See Emma — the AI Assistant.

Sessions

A session is a recorded deposition, interview, or meeting. Sentinel transcribes the audio with speaker diarization (who said what), and the transcript becomes searchable like any other content. During live or ambient sessions, Sentinel can surface AI "leads" — related documents, possible contradictions, and follow‑ups — in real time. See Sessions & Transcription.

Findings

Findings are the notebook of a matter: notes, theories, legal issues, contradictions, and other entries, each optionally linked to the evidence that supports it. Findings are created conversationally — Emma proposes a finding mid‑conversation and you accept, edit, or dismiss it. See Findings.

Practice modes

A tenant runs in exactly one practice mode, which reshapes the navigation and available features:

Mode Objects Distinctive surfaces
Litigation Cases, custodians Discovery & document review, productions, privilege log, motions, jury selection, depositions, communication & issue analytics
Transactional Deals, counterparties Deal data rooms, due‑diligence request lists, deal checklists & templates, business intelligence
Real Estate Properties, stakeholders Deal pipeline, diligence, closing checklists (transactions ribbon)

Practice mode is a per‑tenant setting that an administrator controls (and that Sentinel sets during provisioning). Changing it changes what every user in the tenant sees. A few surfaces are practice‑mode‑specific and are noted in their guides.

Access control: who can see a matter

Sentinel follows an "administrators assign attorneys to matters" model:

  • Admins (and Sentinel's platform‑admin staff role) can access any matter in the tenant.
  • The owner of a personal data‑lake matter can always access it.
  • Enterprise data‑lake matters are org‑wide — every authenticated user in the tenant can access them.
  • Everyone else gets access to a matter only through an explicit assignment made by an admin. Each assignment carries an access level:
    • Full — read and write.
    • Read‑only — can view but not modify.
    • Limited — a restricted subset.

Mutating actions (creating, editing, deleting within a matter) require full access. Assignments can be removed, which immediately revokes access. See Administration → Matter assignments & access levels.

Multi‑tenancy and data isolation

Each customer is a tenant. A tenant's documents live in that tenant's own database and storage — customer data is not commingled across tenants. A central platform layer handles cross‑tenant concerns like accounts and tenant registration, but not customer document content. See Security & Compliance for the full posture.


With these concepts in hand, the User Guides cover each feature in depth. Start with Matters, Cases & Deals.

Matters, Cases & Deals

Matters, Cases & Deals

A matter is the container for a piece of work and everything tied to it. In Litigation mode a matter is a case; in Transactional mode it's a deal; in Real Estate mode it's a property transaction. The concepts below apply to all three; where they differ, it's noted.

The matters list

Open Matters from the sidebar (labeled Deals or Deal Pipeline in transactional / real‑estate modes) to see the matters you have access to. Administrators see all matters in the tenant; other users see the matters they've been assigned to plus any organization‑wide (enterprise data‑lake) matters.

From the list you can open a matter, and — depending on your role — create a new one. Transactional and real‑estate modes also offer a pipeline view that arranges deals by stage.

Selecting your current matter

Use the matter selector in the sidebar to set your current matter. Doing so expands the sidebar to that matter's workspace and updates the URL so you can bookmark or share a deep link. You can switch matters at any time.

Inside a matter

A matter opens to an overview and a set of tabs that organize its work. The exact tabs depend on practice mode and matter type, but generally fall into three groups:

Case / deal information

  • Overview — the matter's header and dashboard: title, number, status, priority, and key facts.
  • Parties — the people and organizations involved (in litigation: plaintiffs, defendants, counsel; in deals: counterparties and advisors).
  • Timeline — a chronology of key events, including events extracted from the matter's documents (see Timelines & Summaries).
  • Court Docket (litigation) — docket entries imported from a court source, where you can find the operative complaint, motions, and orders. See Integrations → Court & docket import.
  • Pleadings (litigation) — the matter's pleadings.

Documents & discovery

  • Data Room / Evidence — the matter's documents and received evidence. See Data Rooms & Uploading Documents.
  • Working Files / Case Files — firm‑authored work product kept separate from discovered evidence.
  • Interviews / Depo Prep (litigation) — interview templates, scheduled sessions, and recordings. See Sessions & Transcription.

Analysis

  • Findings — the matter's notebook of notes, theories, issues, and contradictions. See Findings.
  • Checklists — for deals, the due‑diligence checklist; for litigation, case‑specific checklists and hearing preparation. See Deal Checklists & Request Lists.
  • Hearing / Trial prep (litigation) — preparation surfaces for upcoming proceedings.
  • Team — who is assigned to the matter and at what access level (managed by admins). See Administration.

Creating a matter

How matters are created depends on practice mode and your workflow:

  • New matter intake — litigation tenants can create a matter from an intake form.
  • Import from a court docket — create a litigation case directly from a CourtListener or UniCourt docket search. See Integrations → Court & docket import.
  • New deal — transactional / real‑estate tenants create deals from the deals list or pipeline.

Newly created matters start empty; add documents (see Data Rooms & Uploading Documents) and assign team members (admins, via the Team tab) to begin work.

Matter status and lifecycle

Matters carry a status (e.g. active, closed/archived) and a priority. Closed or archived matters are typically hidden from default lists but remain accessible. Closing a matter does not delete its content.

Tips

  • The fastest way to navigate within a matter is to ask Emma — e.g. "take me to the court docket" or "open document review filtered to John Smith's emails." See Emma.
  • Deep links are stable: copy the URL of any matter view to share it with a colleague who also has access.

Data Rooms & Uploading Documents

Data Rooms & Uploading Documents

The Data Room is a matter's primary document repository — received evidence and discovery documents in litigation, or deal documents in transactional and real‑estate work. Everything you put here is ingested and indexed automatically (see Core Concepts → The content pipeline), so it becomes searchable and available to Emma.

Opening the data room

Select a matter, then open Data Room from the matter's sidebar section. Transactional and real‑estate modes also expose a Data Rooms catalog where you pick a deal to open (or create) its room.

Uploading documents

You can add documents several ways:

  • Drag and drop files or folders directly into the data room.
  • Chunked upload for large files and bulk sets — uploads resume and report progress.
  • Email containers and archives — drop in PST, MBOX, MSG, or ZIP files; Sentinel automatically splits them into their individual messages and attachments during ingestion.
  • Media — audio, image, and video files are supported and run through multimodal extraction so their content is searchable.

After upload, each document is processed asynchronously. You'll see it move through processing states (queued → extracting/OCR → embedding → complete) and appear in search once complete. See Core Concepts → What you see while a document is processing.

Organizing

Within a data room you can organize documents into folders, search and filter them, tag them, and open any document for review. For full review features — coding, bulk actions, email threading — open Document Review.

Client upload portal

You can collect documents from a client or third party without giving them an account. Sentinel mints a secure upload link you send to the client; files they upload land in the matter and flow through the same ingestion pipeline. Uploaded files carry provenance — the originating link, the uploader, and an integrity hash — and are marked as a client upload so reviewers can see where a document came from. The upload's processing status is visible to the inviting attorney so you can confirm when the client's documents have finished indexing.

Sharing a data room with an external guest (read‑only)

You can invite an outside party — opposing counsel, a client, a co‑counsel — to view a data room without a tenant account:

  1. An admin or attorney invites the guest by email. Sentinel generates a time‑limited invite link.
  2. The recipient opens the link and confirms the invited email address at a gate.
  3. They get a short‑lived, read‑only session: a document list and inline preview. Guests can view only — they can't edit, tag, or download beyond what the share allows.

Guest access is re‑checked on every request, so revoking a guest takes effect immediately. Guest sharing is managed from the data room's access controls; see Administration → Data‑room guest sharing.

Emailing documents (a lighter alternative). For a one‑off send where the recipient doesn't need ongoing access, you can generate short‑lived, read‑only download links for selected documents and send them by email. This is separate from inviting a guest into the room.

Productions vs. the data room

The data room holds the documents you work with. When you need to formally hand documents to another party — Bates‑numbered and stamped — you build a production from selected documents. See Productions & eDiscovery.

How uploaded content becomes searchable

Regardless of how a document arrives — manual upload, client portal, docket import, or a monitored mailbox — it goes through the same ingestion pipeline and ends up in the same searchable index. There are no separate, unsearchable stores. This is why a deposition transcript, a produced PDF, and a synced email are all findable the same way and all visible to Emma. See Search & Research.

Document Review

Document Review

Document Review (Doc Review) is the full‑screen workspace for reviewing, coding, and tagging documents in a matter. It's the heart of discovery review in litigation and document diligence in deals.

The three‑pane workspace

Document Review uses a three‑column layout:

  • Left — document list. The set of documents you're reviewing, with search, faceted filters (date range, sender/recipient, custodian, tags, flags, responsiveness, document type), and saved searches. Move through the set one document at a time or work in bulk.
  • Center — the document. A viewer with the rendered document (PDF with a selectable text layer), plus email family and threading views so you can see an email together with its attachments and the rest of its conversation.
  • Right — coding & context. Where you code the document, apply tags, add notes, and see AI suggestions.

Coding and tagging

  • Coding assigns a structured determination to a document — for example a relevance/responsiveness code or privilege call. Coding is how a review produces its work product.
  • Tags are flexible labels (issues, custodians, "hot," etc.) you can apply to organize and later filter documents.
  • Flags (such as marking a document "hot") highlight documents that need attention.

You can apply these to a single document, or in bulk to many documents at once — for example, tagging every email from a custodian within a date range.

Searching and filtering within review

The left pane's filters narrow the working set. You can combine keyword search with facets (sender, custodian, date, tag, flag, type) and save a search to return to it. You can also ask Emma to set the filters for you — e.g. "show me unreviewed documents mentioning the merger from Q2 2022" — and she'll apply them in place. See Emma.

AI assistance during review

Emma can act inside the review workspace: advance to the next document, mark responsiveness or toggle a flag, apply or remove tags (individually or in bulk), and surface coding suggestions. This lets you drive review by voice or chat while keeping your hands on the document.

Predictive coding (TAR / CAL)

For large review sets, Review Intelligence provides predictive coding — Technology‑Assisted Review (TAR) / Continuous Active Learning (CAL):

  1. Create a project, choosing the criteria and tags it should predict and a confidence threshold.
  2. Run it across a document set; the project reports progress (processed, remaining, failed).
  3. Use validation coding to hand‑code a sample so the model refines its predictions.

Predictive coding helps prioritize the most likely‑relevant documents and measure review completeness. Projects show status (draft, running, validating, paused, completed, failed) as they progress.

Analytics

Review Analytics summarize the work: tag usage by category, how many documents are coded, and review progress. Drilling into a tag lists the documents that carry it and jumps you into review pre‑filtered to that tag. See also Relationship Intelligence & Calendar for communication analytics.

Privilege review

Documents identified as privileged or work product can be recorded in the matter's privilege log, with the privilege basis and redaction status, ready to export for a privilege assertion. See Productions & eDiscovery → Privilege log.

Tips

  • Use saved searches to define review batches and return to them.
  • Bulk actions are the fastest way to handle obvious sets (e.g., everything from a single custodian) before fine‑grained review.
  • Everything you code and tag feeds Analytics and is visible to Emma and search.

Search & Research

Search & Research

Sentinel makes everything in a matter searchable through one index, and offers several search modes so you can match the way you're searching to what you're looking for.

Search modes

Mode Best for How it works
Keyword Exact strings, names, Bates numbers, Boolean (AND/OR/NOT, phrases) Full‑text search over extracted text; fast and precise.
Semantic Conceptual / paraphrased questions ("documents about regulatory pressure") Matches by meaning using vector embeddings, even when wording differs.
Hybrid Most natural‑language questions Combines keyword precision with semantic recall — a strong default.
Visual Visual content: signatures, handwriting, diagrams, photos, scanned receipts Searches the multimodal embeddings of images and scanned pages — things plain text search can't find.

Results come back ranked, each with a snippet and a link to open the document. Scores are mode‑relative — a higher score is better within a mode, but scores aren't comparable across modes.

Filters

You can narrow results with filters such as custodian, sender email, and document type, on top of any mode. In the Document Review workspace these become faceted filters you can stack and save.

What gets searched

Search spans a matter's indexed content, including:

  • Data‑room documents and case files (firm work product).
  • Email (including messages and attachments extracted from PST/MBOX/MSG archives), with metadata like sender, recipients, and subject.
  • Session transcripts from depositions, interviews, and meetings.
  • Court docket filings imported into the matter.

Because all of these flow through the same ingestion pipeline, they're searchable the same way and to the same depth.

Research with Emma (Privileged Chat)

The Privileged Chat / Research surface is the unified place to ask questions in natural language and get answers grounded in your documents. It has two layouts:

  • No matter selected — a focused chat with Emma, plus your recent conversations and history.
  • Matter selected — a workspace combining the chat, a document viewer, and a Sources panel. When Emma cites a document, you can open it right there and read the passage she relied on.

Emma searches your evidence for you, summarizes what she finds, and answers with verifiable citations — clickable references that open the underlying document. She will tell you plainly when retrieval finds little or nothing rather than filling the gap with general knowledge. See Emma — the AI Assistant for the full picture, including voice and guardrails.

Conversations

Research conversations are matter‑scoped (or global when no matter is selected). You can start new threads, browse and search your history, and export a conversation. You can also ask Emma to save a chat to the matter's case files so it's retained and searchable later.

Tips

  • Start broad with hybrid search or a question to Emma; switch to keyword when you know the exact term, name, or Bates number.
  • Use visual search when you're looking for something on the page rather than in the text (a signature, a stamp, a diagram).
  • Open the Sources panel while researching so you can verify Emma's answers against the documents as you go.

Emma — the AI Assistant

Emma — the AI Assistant

Emma is Sentinel's built‑in AI assistant. She helps you find documents, answer questions about a matter, analyze communications, navigate the app, code and tag documents, and review sessions — by text or by voice. Emma is powered by Anthropic's Claude (currently Claude Sonnet 4.6) and is grounded in your matter's evidence rather than the open web.

Where to find Emma

  • Privileged Chat / Research — the main place to converse with Emma. With a matter selected, you get a workspace with the chat, a document viewer, and a Sources panel. See Search & Research.
  • The Home omnibox — type a question or command to start with Emma.
  • In context — Emma can act inside the Document Review workspace and reference the matter you're currently in.

What Emma can do

Emma works through a set of tools. Grouped by what they accomplish:

  • Find documents — keyword, semantic, hybrid, and visual search across the matter; find a specific document by filename, email subject, or Bates number; open a document to read it.
  • Answer questions about the matter — summarize the matter, pull the case timeline, surface detected issues, and answer content questions grounded in retrieved documents.
  • Analyze communications — who communicated with whom, email threads and the messages within them, communication patterns, and anomalies.
  • Work the court docket (litigation) — find filings like the operative complaint, motions, and orders by type or query.
  • Review actions — advance documents, mark responsiveness, toggle flags, apply and remove tags (including bulk tagging by criteria), and set review filters in place.
  • Sessions — list a matter's depositions/interviews/meetings, read or filter a transcript, get a session summary, search across all transcripts, and review what the ambient system surfaced during a session.
  • Navigate and teach — take you to a destination in the app, or launch an interactive walkthrough that both navigates and teaches a feature when you ask "how does X work."
  • Capture work — propose a finding for you to accept, and save a chat to the matter's case files so the conversation is retained and searchable later.

There is no separate manual "notes" feature — asking Emma to save the chat is how you retain a conversation for later reference.

Text and voice

  • Text — type questions and commands; Emma replies in clean prose with clickable citations, and can take actions through her tools.
  • Voice — speak to Emma in real time. Voice sessions are conversational (you talk, Emma responds aloud) and are useful hands‑free, such as during review or session prep.

Citations and guardrails

Emma is built for defensible legal work, so her output is constrained:

  • Citation grounding. The only way Emma may reference a document is with a citation token (e.g. [cite:c1]) that a tool actually returned during the conversation. These tokens map to real documents and render as clickable chips that open the document.
  • Mechanical verification. After Emma composes a response, it is checked against the documents that were actually retrieved. If the response references a document, filename, or Bates number that wasn't returned by a tool, the response is blocked rather than shown. This prevents fabricated citations from contaminating work product.
  • No padding from memory. When retrieval returns little or nothing, Emma says so plainly and stops, instead of filling the answer with general knowledge — even for famous public datasets she might "recognize."
  • Matter‑scoped access. Emma can only see matters you have access to. Ask about a matter you aren't assigned to and she can't retrieve it.
  • Per‑conversation budget. Each Emma conversation runs within a cost budget, a guardrail against runaway usage.

Tips for working with Emma

  • Be specific about the matter and the scope: "In the Acme case, find emails between the CFO and outside counsel about the August financing."
  • To reference a document by its name or Bates number, just say so — Emma will look it up by its label rather than searching inside documents.
  • Ask "how does feature work" to get an interactive walkthrough instead of a wall of text.
  • Open the Sources panel to verify Emma's citations as you go.

Related AI features

Emma is the general assistant. Sentinel also has feature‑specific AI:

  • A read‑only Motion Agent that helps with motion work and citation verification — see Motions.
  • Ambient intelligence that surfaces leads live during sessions — see Sessions & Transcription.
  • AI document summaries and timeline extraction — see Timelines & Summaries.

Sessions & Transcription

Sessions & Transcription

A session is a recorded deposition, witness interview, or meeting. Sentinel records it, transcribes it with speaker diarization (identifying who spoke), makes the transcript searchable, and can surface AI assistance live as the session happens.

The Sessions hub

Open Sessions from the sidebar to see a matter's recorded sessions and to start a new one. When creating a session you choose the matter, the participants, and how it will run (for example a live video meeting or an in‑room ambient recording).

Types of session

  • Ambient / in‑room session — captures audio in a room (e.g. a deposition or meeting), with a live transcript, speaker labels, and side panels for objectives, exhibits, and AI suggestions.
  • Live video meeting / interview — a video session that combines conferencing with the same ambient intelligence panels. External participants can join a session through a secure link without an account.
  • Upload a recording — already have audio or video? Upload it to a session for transcription and analysis after the fact.

Transcription and diarization

Sentinel transcribes session audio and assigns speaker labels so the transcript reads as a back‑and‑forth rather than an undifferentiated block. Transcripts are then chunked and embedded like any other content, so you can:

  • Read the transcript, optionally filtered by speaker.
  • Search within and across sessions — find every session where a name, term, or topic came up.
  • Ask Emma to summarize a session or pull what was said on a topic.

Live AI assistance (ambient intelligence)

During a live or ambient session, Sentinel can act as a quiet second chair. As the transcript streams in, the ambient intelligence system watches for signals — possible contradictions with earlier testimony or documents, mentions of key people or topics, and other follow‑ups — and surfaces them as leads in a side panel, with related documents, so the attorney can follow up in the moment.

What a session produces

After a session you have:

  • A transcript with speaker diarization.
  • An AI summary (key points, topics, decisions, action items).
  • Timeline events mentioned in the session, feeding the matter timeline.
  • Any leads the ambient system surfaced and the documents it pointed to.

All of this is searchable and available to Emma alongside the rest of the matter's content.

Exhibits and objectives

In an ambient session you can pull up exhibits (documents from the matter) during the session and keep a list of objectives — the questions or topics you intend to cover — so the session stays on track.

Privacy note

Sessions involve recordings of people. Handle consent and recording in accordance with the law of the relevant jurisdiction and your firm's policies. Transcripts and recordings are stored within your tenant like other content; see Security & Compliance.

Findings

Findings

Findings are a matter's working notebook — the place to capture notes, theories, legal issues, contradictions, and other analysis, each optionally linked to the evidence that supports it. Findings are how the team records what it has concluded from the documents, separate from the documents themselves.

A chat‑first notebook

Findings are created conversationally. As you work with Emma in the matter, she can propose a finding mid‑conversation — surfaced as an inline card. You accept, edit, or dismiss it. Accepted findings collect in a side rail where you can edit, pin, or delete them.

This keeps analysis flowing from the investigation itself: when Emma identifies a contradiction or a key fact while answering a question, capturing it is one click.

Anatomy of a finding

A finding typically has:

  • A title and content (the note, theory, or observation).
  • A type — for example a note, theory, legal issue, contradiction, credibility observation, research note, to‑do, timeline event, or recommendation.
  • Optional linked evidence — the documents that support it, so the finding is traceable back to the record.
  • Optional categories and a pinned flag to keep important findings at the top.

Why link evidence

Because findings can link to the documents that support them, a matter's findings become a navigable map of "what we believe and why." Linked evidence is clickable, taking you straight to the supporting document, and keeps conclusions defensible.

Findings and search

Findings created with linked evidence are part of the matter's analysis. (Note that free‑text notes you write directly, without attaching evidence, are browsable in the Findings tab but aren't part of the document search index — the evidence they link to is.)

Tips

  • Let Emma do the capturing: ask her to note a contradiction or a theory as you uncover it, and accept the proposed finding.
  • Pin the findings that frame the matter (key theory, dispositive issue) so the team sees them first.
  • Use linked evidence liberally — a finding without support is just an opinion; a finding with citations is work product.

Timelines & Summaries

Timelines & Summaries

Sentinel uses AI to turn a matter's documents into two things that make a large record comprehensible: document summaries and an extracted case timeline.

Document summaries

Every document that goes through the ingestion pipeline gets a short, fact‑grounded AI summary — a few sentences describing what the document is and its key points. Summaries appear in search results, document previews, and the review pane, so you can triage a document before opening it.

Summaries are generated automatically during ingestion. They're written to be concise and to stick to what the document actually says.

Case timeline

The timeline is a chronology of events in a matter, including events extracted from the documents themselves. Rather than only showing events you enter by hand, Sentinel can read across the matter's content and assemble an ordered timeline of what happened and when.

How extraction works (at a glance)

Timeline extraction reads across the matter's indexed content, groups related material, and uses AI to identify discrete events — each with a date, participants, a short description, and a risk level (for example: critical, medium, cleared, or neutral) with an explanation. Overlapping events are merged into a single ordered timeline.

You typically run timeline extraction on demand for a matter, and can re‑run it to incorporate newly added documents. For very large matters, extraction can be scoped to a date range to keep events granular.

Using the timeline

  • See the matter's events in chronological order, with risk levels highlighting what matters most.
  • Open the document(s) behind an event.
  • Ask Emma to "show the case timeline" or to focus on a period or topic.

How they fit together

Summaries help you understand individual documents quickly; the timeline helps you understand the matter as a sequence of events. Both are grounded in the matter's actual content and update as new documents are ingested. They feed — and are fed by — your Findings and Search & Research.

AI summaries and extracted events are aids for orientation and triage, not a substitute for reading the underlying documents. Always verify against the record before relying on them.

Deal Checklists & Request Lists

Deal Checklists & Request Lists

For transactional and real‑estate practices, Sentinel provides due‑diligence tooling that tracks what a deal needs and auto‑completes as documents arrive: deal checklists, document request lists, and templates.

These features are specific to transactional and real‑estate practice modes.

Deal checklists

A deal checklist is a structured list of due‑diligence items for a deal, organized into sections (legal, financial, tax, and so on). Each item tracks a status (e.g. pending, in progress, waiting, complete, not applicable, or waived), how it's satisfied, and any documents connected to it.

Templates

Checklists are created from templates — reusable, deal‑type‑specific lists (for example, a purchase & sale checklist or a closing‑variant checklist). Templates are reference data maintained per tenant.

Administrator note: real‑estate checklist templates are seeded into a tenant as a setup step. If the checklist picker is empty or the recommendation card never appears, the templates likely haven't been seeded yet — see your administrator. (Without templates the feature is simply inert; nothing breaks.)

Creating a checklist for a deal

From the deal's Checklists tab, apply a template to create the checklist. Sentinel can also recommend the most appropriate template for a deal: it ranks templates against the deal and its documents and surfaces a recommendation card while the deal has no checklist yet.

Auto‑match: checklists that complete themselves

The standout feature is auto‑match. Sentinel compares the deal's data‑room documents to the checklist items and, where it finds a match, generates a plain‑English rationale and an AI verdict about whether the document actually satisfies the item:

  • The AI verdict (satisfied / partial / unsatisfied) is what gates completion — so a green check never contradicts its own rationale.
  • A keyword match alone is only a prefilter; the verdict decides.

The result is a checklist that fills itself in as diligence documents land, with a visible reason for every completion you can review and override.

Working a checklist

On the checklist detail page you can:

  • Filter items by Outstanding / Complete / All.
  • Mark items complete, waived, or not applicable.
  • Add notes and attach documents to an item.
  • Read the AI rationale inline for auto‑matched items.

Request lists (document request lists)

A request list is a list of documents you're requesting for a deal — diligence requests organized by requesting/responding party and due date. Use request lists to track what's been asked for and what's outstanding, alongside the checklist that tracks what's been satisfied.

Putting it together

A typical diligence flow:

  1. Apply (or accept the recommended) template to create the deal's checklist.
  2. Issue request lists for the documents you need.
  3. As documents land in the data room, auto‑match completes checklist items with rationales.
  4. Review the outstanding items and the AI verdicts, overriding where your judgment differs.

Motions

Motions

The Motions workspace (litigation) is where you work a motion end to end: bring in the relevant brief, extract and verify its citations, organize the authorities, and assemble a response — with an AI Motion Agent that helps and keeps the citation record honest.

Motions is a litigation‑mode feature. It is an actively developing workspace: some steps are fully shipped while others are being built out, and the workspace shows each step's status as you go.

The motions list

From a litigation matter, open Motions to see the motions for that matter, each with its citation count, status, and date. Open one to enter its workspace, or start a new motion.

The motion workspace

The workspace is organized as a sequence of steps:

  1. Upload / attach the brief — bring in the opposing brief (or the brief you're responding to). The brief is frequently already a document in the matter; the Motion Agent will point that out and offer to attach it.
  2. Pull list — the list of citations extracted from the brief.
  3. Library — the authorities relevant to the motion, organized for the response.
  4. Verify — each citation is checked (see below).
  5. Outline — an organized response outline keyed to the authorities.

Citation verification

Sentinel verifies the citations in a motion against legal sources and records the result, so you don't file on a case that doesn't say what the brief claims. Each citation gets a verdict evaluated across multiple dimensions (jurisdiction, recency, how on‑point it is, strength, and so on), with a short explanation and a status indicating whether it checks out, is uncertain, or is a problem.

The set of verifications forms a citation ledger — a gate that tracks which citations are confirmed versus blocking, and whether the motion is ready to file. A printable certificate records the verification status of a motion's citations.

The Motion Agent

The motion workspace has its own AI assistant, the Motion Agent — a litigation‑associate persona focused on the single motion in front of you. It is read‑only with respect to the record: it explains and plans, but it does not silently change verifications, draft and bind authorities, mark a motion complete, or trigger paid research spend on its own. Instead it:

  • Orients itself in the motion (what's attached, the citation count, the verification breakdown) and searches the matter for the relevant brief and authorities.
  • Explains the current plan, the citation ledger, and any individual citation's verdict.
  • Proposes a work plan for you to approve; the actual retrieval, verification, and drafting happen in a run you approve — not unilaterally.

This keeps an attorney in control of every step that affects the filing while still giving you an AI second chair.

Tips

  • Let the Motion Agent orient first — it will often find the opposing brief already in the matter and offer to attach it, saving you an upload.
  • Treat the citation ledger as your pre‑filing checklist: clear the blocking citations before you rely on the motion.
  • The verification certificate is a record of what was checked — useful for your own diligence and for the file.

Jury Selection

Jury Selection

Jury Selection (litigation) is a courtroom workspace for voir dire — managing the venire, capturing impressions, planning strikes with defensible rationale, and arranging the panel. It's a full‑screen surface built for use during jury selection.

Jury Selection is a litigation‑mode feature.

Building the panel

  • Enter jurors individually, or bulk‑import a venire from a CSV.
  • Open a juror detail view to record what you know and learn about each prospective juror.
  • An audit log records the actions taken during selection.

Capturing voir dire

For each juror you can capture impressions during questioning — including by voice at the counsel table — and record attitudes and experiences relevant to your case. The detail view is where your read on each juror lives.

Planning strikes

Jury Selection helps you plan and track strikes:

  • For‑cause and peremptory strikes are tracked per juror.
  • Peremptory strikes prompt you to record a neutral rationale (an attitude/experience basis) so the strike is defensible — supporting a Batson‑type challenge record.
  • The workspace helps you keep your strike plan organized as selection proceeds.

Seating the panel

A seating chart lets you arrange jurors into the box (drag and drop) so you can see the panel as it forms and plan around your remaining strikes.

During live selection

Because selection is fast and high‑stakes, the workspace is designed to stay out of your way — full‑screen, quick entry, and guards so that an in‑progress voice capture isn't lost to an accidental navigation.

Tips

  • Import the venire by CSV ahead of time so you're entering impressions, not names, in the courtroom.
  • Record a concrete, case‑relevant rationale for every peremptory strike as you make it — reconstructing it later is harder and less defensible.
  • Use the seating chart to keep the forming panel and your remaining strikes in view.

Productions & eDiscovery

Productions & eDiscovery

This guide covers the discovery‑production side of litigation: building productions you hand to other parties, tracking received productions, maintaining a privilege log, and bulk‑importing documents through eDiscovery intake.

These are litigation‑mode features.

Productions (outbound)

A production is a formal set of documents handed to another party, with Bates numbering and stamping. In the Productions workspace you:

  1. Select the documents to produce.
  2. Choose a numbering scheme (Bates prefix and padding) and export format.
  3. Generate the production — Sentinel stamps the documents and tracks progress.
  4. Download the export and keep a record of what was produced.

Bates defaults (prefix, padding) can be set tenant‑wide in tenant settings; you can override per production.

Received productions (inbound)

When you receive a production from an opposing party, track it under Received Productions. You can review opposing documents in the same review workspace (filtered to the opposing set) and tie received items back to your document requests.

RFP coverage

RFP Coverage links received production items to your requests for production and tracks which requests have been satisfied — so you can see, at a glance, what the other side has and hasn't produced against each request.

Privilege log

The Privilege Log records documents withheld or redacted on privilege grounds. Each entry captures the document's metadata, the privilege type (e.g. attorney‑client, work product), the basis, whether it's a full or partial redaction, and review status. The log is a searchable, sortable table you can filter and export (for example to serve a privilege log on opposing counsel).

Documents are typically added to the privilege log during Document Review as reviewers make privilege calls.

Review Intelligence (TAR / CAL)

For large collections, predictive coding prioritizes and measures review. See Document Review → Predictive coding.

eDiscovery intake (bulk ingestion)

eDiscovery Intake is the path for bringing large volumes of documents into a matter — bulk uploads and email archives that flow through the same ingestion pipeline (parse → split → OCR/extract → embed → complete) as everything else, so the collection becomes searchable and reviewable. See Core Concepts → The content pipeline.

A typical discovery flow

  1. Collect custodian data into the matter (uploads, mailbox sync, eDiscovery intake).
  2. Review and code documents, making responsiveness and privilege calls.
  3. Log privilege for withheld/redacted documents.
  4. Produce responsive, non‑privileged documents with Bates stamping.
  5. Track received productions from the other side and map them to your RFPs.

Tips

  • Set tenant‑wide Bates defaults once so productions are consistent across the team.
  • Use saved searches and bulk coding to define and clear review batches before producing.
  • Keep the privilege log current during review rather than reconstructing it at production time.

Relationship Intelligence & Calendar

Relationship Intelligence & Calendar

These two features help you stay on top of the people and schedule around a matter: Relationship Intelligence monitors the contacts that matter and surfaces relevant news, and the Calendar keeps your events together with a briefing on who you're meeting.

Relationship Intelligence

Relationship Intelligence tracks the contacts connected to your work — opposing counsel, executives, counterparties, advisors — and aggregates relevant alerts (such as news) about them. It works across matters.

Alerts are grouped by how much attention they need (for example: needs attention, opportunities, market intel, and caught up), each showing its source, headline, sentiment, category, and a relevance signal. You can mark alerts as read, acted‑on, or dismissed, and filter what you see.

A Contacts view holds the underlying relationship database you can browse and maintain.

Communication analytics

Related to relationships, communication analytics analyze the communications within a matter — who talked to whom, email threads and patterns, and anomalies. These help you understand the human network in the evidence and spot unusual activity. Emma can answer the same questions conversationally (see Emma).

Calendar

The Calendar brings your matter‑related events into one place — interviews, sessions, depositions, and manual events — and can sync with an external calendar provider (such as Google Calendar or Microsoft 365) when enabled.

A distinctive touch is the relationship briefing: when you look at an event's attendees, Sentinel enriches each attendee with what it knows from Relationship Intelligence — their company, title, recent news, and last outreach — so you walk into a meeting prepared.

You can create events scoped to a matter and add attendees. Calendar sync settings control which provider (if any) to connect.

Tips

  • Use Relationship Intelligence to keep tabs on opposing counsel and key players without manually trawling the news.
  • Connect your calendar so the attendee briefing can do its job before each meeting.
  • Ask Emma about communication patterns ("who did the CFO email most around the closing?") rather than building the analysis by hand.

Knowledge Base

Knowledge Base

The Knowledge Base is a searchable reference of regulations and compliance frameworks you can consult while working a matter — useful for compliance monitoring and investigations.

What's in it

The knowledge base holds regulations with their title, description, and citation, and can be organized by violation type. Administrators can seed it with common frameworks (for example FCPA and SOX) so the team has a consistent reference.

Using it

  • Search by title, description, or citation.
  • Filter by violation type to narrow to a category.
  • Open an entry to read its detail.

How it fits with the rest of Sentinel

The knowledge base is reference material — the regulatory backdrop against which compliance work happens. It complements:

  • Emma, who brings broad knowledge of federal compliance regimes (FCPA, antitrust, SOX, sanctions/OFAC, BSA/AML, and others) into her analysis of your matter's documents. See Emma.
  • Findings, where conclusions about potential issues are recorded against the evidence. See Findings.

Tips

  • Seed the frameworks relevant to your practice so search returns the citations your team actually uses.
  • Use the knowledge base for the rule, and Emma plus search for how the rule maps onto your matter's facts.

Administration

Administration

This page covers the tenant administration features — managing users and access, API keys, settings, the audit log, and guest sharing. Most of these are admin‑only and appear under the Administration section of the sidebar.

Users and roles

Administrators manage the tenant's users: creating accounts, assigning roles, and handling account recovery.

Roles

Role What it grants
Admin Full tenant administration; access to every matter in the tenant.
Attorney Standard team member; access to assigned and org‑wide matters. The default role.
Deal Team Transactional / M&A focus on deals and diligence.
Platform Admin A Sentinel‑staff role for cross‑tenant operations — not assigned to customer users.

Account actions

From user management an admin can:

  • Create a user (email, name, role). New users may be sent an invitation and a temporary password, and are prompted to set a new password (and enroll in 2FA, if required) on first sign‑in.
  • Reset a password — force a user to set a new password at next login.
  • Reset 2FA — clear a user's two‑factor enrollment so they can re‑enroll (account recovery when a device is lost).
  • Remove a user — deactivate an account so it can no longer sign in.

Matter assignments and access levels

Sentinel's access model is "admins assign attorneys to matters." A user who isn't an admin sees a matter only if:

  • they're assigned to it,
  • they own it (a personal data‑lake matter), or
  • it's an enterprise data‑lake matter (org‑wide).

Each assignment carries an access level:

Access level Can view Can modify
Full Yes Yes
Read‑only Yes No
Limited A restricted subset No

Admins manage assignments from a matter's Team view. Removing an assignment revokes access immediately. Mutating actions within a matter require full access.

Rollout note for admins: when matter‑level access controls are first applied to a populated matter, users without an assignment lose access until assigned. Audit assignments before tightening access so the right people keep the access they need.

API keys

For programmatic and service‑to‑service access, the tenant supports long‑lived API keys (in addition to browser sessions). Manage them from the API Keys admin page:

  • Create a key with a name, an inherited role (no higher than your own), and an optional expiry. The full key is shown once at creation — copy it then; it can't be retrieved later.
  • List keys (admins see all; others see their own), identified by a short prefix.
  • Revoke a key at any time; revocation is immediate.

API‑key management itself can only be done with a browser (JWT) session — an API key cannot create or revoke other keys. See the API Reference → Authentication for the technical details.

Tenant settings

The Tenant Settings page controls tenant‑wide configuration:

  • Practice mode — Litigation, Transactional, or Real Estate (the setting that reshapes navigation and features). This is typically set by Sentinel during provisioning and managed centrally.
  • Branding — display name, logo, and default theme.
  • Production defaults — Bates prefix and padding, default rendition format.
  • Feature toggles — enable/disable features such as Emma and voice mode where applicable.

Audit log

The Audit Log is an admin view of tenant activity for compliance and defensibility. It records actions such as sign‑in successes and failures, sign‑ outs, support sessions, and document/matter operations — with the user, a timestamp, and context (IP and user agent where applicable). The log is built to be reliable: recording an event never blocks the underlying action.

AI‑assistant actions through the API / connector are also audited at the tool‑call level — what was called, by whom, with what arguments, and how long it took — which supports the "what did the assistant see, when, on whose behalf?" question.

Two‑factor authentication (2FA)

Tenants can require 2FA. When required, users enroll an authenticator app on first sign‑in and enter a code at each login. Admins can reset a user's 2FA for recovery. (2FA is enforced at the account/login layer.)

Data‑room guest sharing

Admins and attorneys can invite external, read‑only guests to a data room without giving them accounts, and manage or revoke that access. Guest sessions re‑check access on every request, so revocation is immediate. See Data Rooms → Sharing a data room with an external guest.

Usage

The Usage view shows consumption — for example AI usage and related cost — so admins can monitor spend.

Related

  • Integrations — connecting mailboxes, court sources, and the AI connector.
  • API Reference — auth, REST endpoints, and the connector tool catalog.
  • Security & Compliance — isolation, data residency, and controls.

Integrations

Integrations

Sentinel connects to outside systems to bring data in and to extend the platform. This page covers mailbox monitoring, court/docket import, the AI connector, and bug reporting.

Microsoft 365 mailbox sync (Sentinel Sync)

Sentinel Sync monitors Microsoft 365 mailboxes and streams their messages and attachments into a matter, where they flow through the normal ingestion pipeline and become searchable.

How it works for an administrator:

  1. From the Email Monitoring admin surface, choose a matter and start the authorization flow for a mailbox.
  2. The mailbox owner signs in with Microsoft and grants read access.
  3. From then on, new mail is ingested in the background — deduplicated and indexed alongside the rest of the matter's content.

Sentinel Sync runs as a dedicated service with its own secure, per‑tenant credentials; messages are deduplicated so the same email isn't ingested twice. Connecting or disconnecting a monitored mailbox is an administrative action.

Court and docket import

Sentinel can create litigation cases and pull docket entries directly from court data sources, so you can start a matter from a real docket instead of entering it by hand.

Two providers back a single "Add court dockets" experience:

  • CourtListener — free coverage of U.S. federal (RECAP) dockets; always available.
  • UniCourt — broader coverage (including state courts and document retrieval) when credentials are configured for the tenant.

From the Import Court Cases surface you search for a case, then create the matter (and optionally pull documents). Imported docket entries appear on the matter's Court Docket tab, where Emma can help you find specific filings (e.g. "the operative complaint" or "any motions to dismiss"). See Matters → Court Docket.

If the court‑import option doesn't appear, the provider may not be configured for your tenant.

The Sentinel connector (Model Context Protocol)

Sentinel exposes a secure connector that lets an external AI client — such as an enterprise Claude deployment — work with a matter's documents through a defined set of tools (search, retrieve, summarize, verify citations, tag/code, create findings, and more). The connector speaks the Model Context Protocol (MCP), authenticates with an API key, enforces the same matter‑access rules as the app, and audits every tool call.

This is how you connect Sentinel to a compatible AI assistant outside the product while keeping access controlled and logged. For the full tool catalog and auth details, see the API Reference.

Bug reporting (Report a problem)

Users can report a problem from within the app — by button or by asking Emma — which files a structured report (including context such as the matter and a screenshot) to Sentinel's issue tracker. This gives the team the detail needed to diagnose an issue quickly. The feature requires a per‑tenant configuration to be enabled.

Email sending

Transactional emails Sentinel sends — such as account invitations, credentials, and notifications — go out through an email delivery service. This is platform infrastructure rather than something you configure per matter.

Related

  • Administration — managing the people and keys behind these integrations.
  • Security & Compliance → Subprocessors — the third parties involved in these integrations.

API Reference

API Reference

Sentinel exposes a tenant REST API and a secure AI connector (Model Context Protocol). The connector is the supported way for an external AI client to work with a matter's documents; the REST API primarily serves Sentinel's own web app and service clients.

This reference covers authentication, API‑key management, the public health endpoints, the connector tool catalog, and a representative set of REST endpoints.

Base URL. In production, Sentinel's web UI (a static site) and its API are on different hosts. Call the tenant API host, not the web‑app host. Your tenant's API base URL is available from your administrator. All API paths below are under /api.

Authentication

Every /api/* request must be authenticated. Two schemes are supported.

Bearer JWT (browser sessions)

Authorization: Bearer <jwt>

Short‑lived (15‑minute) JSON Web Tokens issued by Sentinel's control plane when a user signs in. Used by the web app. These carry the user's identity, role, and tenant.

API key (service‑to‑service)

Authorization: ApiKey sk_live_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

Long‑lived, revocable keys for programmatic and service clients. Key facts:

  • Format: sk_live_ followed by 32 hex characters (40 characters total, 128 bits of entropy).
  • Only a SHA‑256 hash of the key is stored. The raw key is shown once at creation and is never recoverable afterward.
  • A short prefix (e.g. sk_live_a1b2) is stored for display so you can identify a key without seeing the secret.
  • Keys are revocable per‑row; revocation takes effect immediately. Keys may carry an optional expiry.
  • A key inherits a role (no higher than the creator's), and that role bounds what the key can do.

For OAuth‑based clients (such as an enterprise Claude marketplace connector), an OAuth client_credentials flow issues an API key that is then presented as a bearer token (Authorization: Bearer sk_live_...).

Key management is session‑only

API‑key management endpoints accept only a Bearer JWT — an API key cannot create or revoke keys. This prevents a leaked key from minting more keys.

API‑key management endpoints

Method Path Auth Purpose
POST /api/api-keys JWT only Create a key (name, optional role ≤ caller's, optional expiresInDays). Returns the raw key once.
GET /api/api-keys JWT only List keys (admins: all tenant keys; others: own). Returns public fields only — never raw keys.
DELETE /api/api-keys/:id JWT only Revoke a key (owner or admin). Idempotent.

Creating, listing, and revoking keys is also available in the app at the API Keys admin page (see Administration → API keys).

Public endpoints (no auth)

Method Path Returns
GET /api/version { service, commit, buildTime } — the running build.
GET /api/health Health status (including DB), plus the same build identifiers.

These are handy for verifying exactly what's deployed.

The Sentinel connector (MCP)

The connector lets a compatible AI client operate on a matter through a defined tool set. It speaks the Model Context Protocol, so the client discovers each tool's input schema automatically. Key properties:

  • Authentication — the connector authenticates with an API key (above).
  • Access control — every tool enforces the same matter‑access rules as the app. Read tools require read access to the matter; write tools require full access.
  • Auditing — every tool call writes an audit record: which tool, by whom, with what arguments, the result size, the duration, and any error.

Read tools

Tool Purpose
list_matters List the matters the caller can access (cases and deals).
get_matter Get a single matter's details (title, status, priority, type, description).
get_matter_summary An AI executive summary of a matter.
search_documents Search a matter's documents. mode is required: keyword, semantic, hybrid, or visual. Optional filters: custodianName, senderEmail, documentType. Returns ranked hits with snippets.
get_document_text Full extracted text + metadata for a document, with a viewer link.
get_case_timeline The matter's chronology of events with risk levels.
get_focus_issues High‑priority issues/contradictions tracked on the matter.
get_detected_issues AI‑detected risks and inconsistencies.
get_privilege_log Privileged / work‑product documents recorded for the matter.
get_email_threads Email conversations grouped by thread.
get_thread_messages Messages within a single email thread.
list_recorded_statements Recorded video/audio statements (e.g. interviews) for a matter.
get_recorded_statement A single recorded statement's transcript and metadata.
search_recorded_statements Search across recorded‑statement transcripts.
verify_citations Validate document references in a draft against the matter's documents.

Write tools

Write tools require full access to the matter.

Tool Purpose
tag_documents Add or remove tags on a set of documents.
tag_documents_by_query Tag documents matching a search query.
code_documents Assign or clear a review code on a set of documents.
code_documents_by_query Code documents matching a search query.
create_finding Create a notebook entry (note, theory, legal issue, contradiction, etc.) with optional linked evidence.
add_to_focus_issue Append analysis to an existing focus issue.

The connector's tool set (used by external clients) is related to but distinct from Emma's in‑app tools (which also include navigation and review‑UI actions). See Emma.

Representative REST endpoints

Beyond the endpoints above, the tenant API backs the web app across matters, documents, search, productions, administration, and integrations. A representative selection:

Area Example endpoints
Matters GET /api/matters, GET /api/matters/:matterId, document sub‑routes under a matter
Tenant settings GET /api/tenant-settings, PUT /api/tenant-settings (admin)
Users & access GET/POST /api/users, password/2FA reset, matter team assignment (admin)
Audit GET /api/audit-logs (admin)
Data‑room sharing guest‑share verification and email‑link endpoints
Court import GET /api/court-import/providers, GET /api/court-import/search, bulk‑create
Mailbox sync GET /api/sentinel-sync/status, monitored‑user authorization

These first‑party endpoints exist primarily to serve Sentinel's own client and may evolve. For a stable, supported integration, prefer the connector and the API‑key auth above. If you need a documented contract for a specific REST endpoint, contact Sentinel.

Conventions

  • Access denials. Routes keyed on a single resource ID return 404 when access is denied (rather than 403), so resource IDs can't be enumerated. List endpoints filter results to what you can access rather than failing the whole call.
  • Roles. Your role (and, for keys, the key's inherited role) bounds what you can do; many admin endpoints require the admin role.
  • Write access. Mutating a matter requires full access to that matter.

Security & Compliance

Security & Compliance

Sentinel handles privileged and confidential legal data, so security and data handling are central to the product's design. This page describes the posture in plain terms. It reflects how the platform is built; for contractual specifics (your DPA, the current subprocessor list, and any certifications), contact Sentinel.

Multi‑tenant isolation

Sentinel is multi‑tenant, but customer data is not commingled:

  • Each customer is a tenant with its own database and its own storage for documents. One tenant's content is not stored alongside another's.
  • A separate platform layer handles cross‑tenant concerns (accounts, tenant registration) — it does not hold customer document content.
  • Background processing (text extraction, OCR, embeddings) is tenant‑scoped: each unit of work carries its tenant context so it's processed against the correct tenant's resources.

Data residency and hosting

Sentinel runs on Microsoft Azure. A tenant's application, database, and document storage live in Azure resources dedicated to that tenant. This keeps customer content within a known, controlled cloud environment rather than spread across ad‑hoc third‑party stores.

For customers requiring isolated network connectivity into their environment, Sentinel can deploy a private network bridge (an overlay edge router) so traffic doesn't traverse the public internet.

Encryption

  • In transit — all client‑to‑server and server‑to‑service traffic uses HTTPS/TLS. Session cookies are HTTP‑only and marked secure in production.
  • At rest — databases and storage use the cloud provider's managed encryption.

How AI providers handle your data

Sentinel uses AI models for inference (Emma, embeddings, extraction, transcription, summaries). The data‑handling posture:

  • Document embeddings and multimodal extraction run through Google Cloud Vertex AI configured for Zero Data Retention (ZDR) — your content isn't retained by the provider after the call and isn't used to train models. (One real‑time voice path is an exception noted in our internal documentation; it is not used for storing document content.)
  • Assistant inference (Emma) uses Anthropic's Claude. API usage under commercial terms is not used to train models.
  • Transcription and certain text inference use Azure AI services, isolated to Sentinel's subscription and not used for training.

The practical summary: customer content is stored and managed within the tenant's Azure resources, and where content is sent to a model provider for inference, it's over secure channels under terms that don't retain it for training.

Access control

Access is enforced at multiple layers:

  • Authentication — every API request requires a valid session (a short‑lived JWT) or a revocable API key. See API Reference → Authentication.
  • Roles — admin, attorney, deal team, and a Sentinel‑staff platform role bound what a user can do.
  • Per‑matter access — non‑admins access a matter only via an explicit assignment (or ownership / org‑wide data lakes), with an access level of full, read‑only, or limited. Mutations require full access. See Core Concepts → Access control.
  • Resource‑ID safety — single‑resource routes return 404 on a denied request so IDs can't be enumerated.
  • Two‑factor authentication — tenants can require 2FA for all users.

Audit logging and defensibility

  • Activity audit — sign‑ins, sign‑outs, support sessions, and document/matter actions are recorded with user, timestamp, and context. Recording an event never blocks the underlying action.
  • AI tool‑call audit — every action an AI assistant takes through the connector is logged at the tool‑call level (which tool, by whom, with what arguments, how long it took, any error), answering "what did the assistant see, when, on whose behalf?"
  • Citation grounding — Emma's responses are mechanically verified so she can't reference documents that weren't actually retrieved, keeping AI output defensible as work product. See Emma → Citations and guardrails.

Subprocessors

Third parties that may process customer data, by category (the authoritative, current list is maintained by Sentinel and provided with your DPA):

Category Provider
Cloud hosting, storage, database, OCR, speech, some inference Microsoft Azure
AI assistant inference Anthropic (Claude)
Embeddings & multimodal extraction (ZDR) Google Cloud (Vertex AI / Gemini)
Transactional email delivery Email delivery provider
In‑app issue / bug reporting Issue‑tracking provider

Additional providers may apply to optional features (for example telephony or payments) only when those features are enabled.

Compliance direction

Sentinel is built toward enterprise compliance expectations — segregation of duties in its release process, comprehensive audit logging, least‑privilege access, and tenant isolation. For the current status of formal attestations (such as SOC 2) and your specific contractual commitments, contact Sentinel.

This page is a plain‑language description of the platform's security design, not a contract or a certification. Authoritative security commitments are made in your agreement and supporting documentation.

FAQ

Frequently Asked Questions

Getting started

What is Sentinel?

Sentinel is an enterprise platform for legal compliance monitoring and eDiscovery. It ingests documents and communications, makes them searchable, provides a review workspace, and offers an AI assistant (Emma) grounded in your matter's evidence. See Introduction.

How do I sign in?

Open your tenant's Sentinel URL and sign in with your email and password. New accounts may be prompted to set a new password and enroll in two‑factor authentication on first login. See Getting Started.

I forgot my password / lost my 2FA device.

Use the Forgot password link to reset your password. For a lost 2FA device, ask a tenant administrator to reset your 2FA so you can re‑enroll.

Why does the navigation look different from a colleague's?

Two reasons: your tenant's practice mode (Litigation, Transactional, or Real Estate) changes whole sections of the app, and your role controls whether you see administration tools. See Core Concepts → Practice modes.

Documents & search

What file types can I upload?

Documents (PDF, Office files), images, audio, and video, plus email containers (PST, MBOX, MSG) and archives (ZIP), which Sentinel splits into their individual items. See Data Rooms & Uploading Documents.

I uploaded a document but can't find it in search yet.

Ingestion is asynchronous. A document moves through processing states (queued → extracting/OCR → embedding → complete) and becomes searchable once complete. Large files and media take longer. See Core Concepts → What you see while a document is processing.

What's the difference between the search modes?

Keyword matches exact terms; semantic matches by meaning; hybrid combines both; visual searches the content of images and scanned pages. See Search & Research.

Can Sentinel read scanned documents and images?

Yes. Scanned PDFs and images run through OCR, and images/audio/video go through multimodal extraction, so their content is searchable.

Emma (the AI assistant)

Can I trust Emma's citations?

Emma may only cite documents that a tool actually returned, and her responses are mechanically verified — a response that references an unretrieved document is blocked. She also says so plainly when retrieval finds little, instead of padding from general knowledge. See Emma → Citations and guardrails.

Does Emma see all my matters?

No. Emma can only access matters you have access to.

Can I talk to Emma instead of typing?

Yes — Emma supports real‑time voice in addition to text.

Does using Emma train an AI model on my data?

No. Inference is performed under commercial terms that don't use your content for training, and embedding/extraction runs with zero data retention. See Security & Compliance → How AI providers handle your data.

Administration

How do users get access to a matter?

An administrator assigns them, with an access level of full, read‑only, or limited. Admins can access any matter; some matters are org‑wide. See Administration → Matter assignments & access levels.

How do I add or remove a user?

From user management, an admin can create users, reset passwords and 2FA, and deactivate accounts. See Administration → Users and roles.

What are API keys for?

Long‑lived, revocable keys for programmatic/service access — including the AI connector. Manage them on the API Keys admin page; the raw key is shown only once. See Administration → API keys and the API Reference.

Can I change our practice mode?

Practice mode is a tenant‑wide setting managed centrally (set during provisioning). Talk to Sentinel or your administrator if it needs to change — it reshapes the experience for everyone in the tenant.

Integrations

Can Sentinel pull in email automatically?

Yes — Sentinel Sync monitors Microsoft 365 mailboxes and ingests their mail into a matter. See Integrations → Microsoft 365 mailbox sync.

Can I start a case from a court docket?

Yes — import from CourtListener (free, U.S. federal) or UniCourt (broader, when configured). See Integrations → Court and docket import.

How do I report a bug?

Use the in‑app "Report a problem" option, or ask Emma. It files a structured report (with context) to Sentinel's issue tracker.

Security

Where does our data live?

In your tenant's own database and storage on Microsoft Azure, isolated from other tenants. See Security & Compliance.

Is our data used to train AI models?

No — see the AI data‑handling section in Security & Compliance.

Is everything logged?

Key activity and every AI tool call are audited for compliance and defensibility. See Administration → Audit log.

Is Sentinel SOC 2 certified?

For the current status of formal attestations and your specific contractual commitments, contact Sentinel. See Security & Compliance → Compliance direction.

Exporting this documentation

Can I get this as a PDF?

Yes. Open the single printable version and use your browser's Print → Save as PDF, or print any individual page (its sidebar and nav are hidden in print).

Glossary

Glossary

Key terms used throughout Sentinel and this documentation.

Ambient intelligence — AI that watches a live or recorded session's transcript and surfaces leads (related documents, possible contradictions, follow‑ups) in real time. See Sessions & Transcription.

Bates number — a sequential identifier stamped on produced documents for unambiguous reference. Set via a prefix and padding when building a production.

Case — a litigation matter. See Matter.

Case files — firm‑authored work product (drafts, research, pleadings) kept distinct from discovered evidence in a matter.

Citation grounding / citation token — Emma may only reference a document via a citation token (e.g. [cite:c1]) that a tool actually returned; responses are verified against those tokens so fabricated references are blocked. See Emma.

Content pipeline — the multi‑stage ingestion process (parse → split → OCR/extract → embed → complete) every file goes through to become searchable. See Core Concepts.

Custodian — a person whose documents/communications are collected in discovery.

Data room — a matter's primary document repository. In Sentinel, the matter is its data room. See Data Rooms.

Deal — a transactional or real‑estate matter. See Matter.

Deal checklist — a template‑driven due‑diligence checklist that can auto‑complete as documents arrive. See Deal Checklists.

Diarization — labeling a transcript by speaker (who said what). See Sessions.

Embedding — a vector representation of content's meaning that powers semantic and visual search.

Emma — Sentinel's built‑in AI assistant. See Emma.

Enterprise data lake — an organization‑wide matter every tenant user can access.

Finding — a notebook entry (note, theory, legal issue, contradiction, etc.), optionally linked to supporting evidence. See Findings.

Guest (data‑room guest) — an external party invited to view a data room read‑only, without a tenant account.

Hybrid search — search combining keyword precision and semantic recall.

Matter — the top‑level container for a piece of work and everything related to it. A case (litigation), deal (transactional), or property transaction (real estate). See Matters, Cases & Deals.

Matter assignment — an admin granting a user access to a matter at an access level (full, read‑only, or limited).

MCP (Model Context Protocol) — the protocol Sentinel's AI connector speaks so external AI clients can use its tools. See API Reference.

OCR (optical character recognition) — recovering text from scanned PDFs and images during ingestion.

Personal data lake — a personal matter owned by a single user.

Practice mode — the per‑tenant setting (Litigation, Transactional, or Real Estate) that reshapes navigation and features. See Core Concepts.

Privilege log — a record of documents withheld or redacted on privilege grounds, with basis and status. See Productions & eDiscovery.

Production — a formal, Bates‑stamped set of documents handed to another party. See Productions & eDiscovery.

Request list (DRL) — a list of documents requested for a deal, tracked by party and due date. See Deal Checklists & Request Lists.

Semantic search — meaning‑based search using embeddings.

Session — a recorded deposition, interview, or meeting that Sentinel transcribes and analyzes. See Sessions.

Sentinel Sync — the service that monitors Microsoft 365 mailboxes and ingests their mail. See Integrations.

TAR / CAL — Technology‑Assisted Review / Continuous Active Learning; predictive coding for large review sets. See Document Review.

Tenant — a single customer organization on Sentinel, with isolated data.

Visual search — search over the visual content of images and scanned pages (signatures, diagrams, photos) via multimodal embeddings.

Zero Data Retention (ZDR) — a configuration where an AI provider doesn't retain submitted content after processing. See Security & Compliance.