The eDiscovery MCP server for Claude and ChatGPT.
Claira MCP connects the AI assistant your team already uses to Claira. Ask it to check a bulk review, run privilege and relevance review or pull cited insights, and your documents stay in Claira. Early access is open to a small group of teams.
Everything you’d do in Claira, asked for in plain words.
The capabilities are tools your assistant calls for you. There is nothing new to learn and no scripting.
Ask for status, progress, failures and skipped documents, and what to do about each.
Describe the need. Claira drafts the review from your case context, tests it on sample documents and shows the scope before anything runs.
Ask which documents came back privileged, responsive, hot or unsure, filtered and ready to act on.
Chronologies, narratives, relevance summaries and PII reports, with the source documents they draw on.
Paste in the amended complaint and review a section-by-section update to the case context every review is built from.
Copy and map values between review fields before a production, with a compatibility check and a one-document test first.
Works with the assistant your team already uses.
Claira MCP is a standard remote MCP server. Add it once as a connector and it works wherever the Model Context Protocol does.
Claude Desktop, claude.ai and Claude mobile. Add Claira as a remote connector, sign in, and run review from the conversation.
Add Claira as a custom connector in ChatGPT for the same review tools and the same confirmations on our side.
It speaks the open Model Context Protocol over Streamable HTTP with OAuth, so any compliant assistant can connect.
Review results, next to the rest of the work.
An API connects software to software. An MCP server connects your evidence to the assistant where the rest of the legal work happens. The same assistant that reads your Claira results drafts the correspondence, so the chronology becomes the brief section, the results table becomes the client update, and the skipped-document list becomes tomorrow’s plan.
The questions your GC will ask.
By default your assistant sees results, statuses and citations, never document text. Claira’s own models, in your residency region, are the only AI that reads your documents.
Organization admins enable the MCP server and choose what it may return for each case.
Anything that uses tokens or writes to a review field shows its scope first and waits for your confirmation. Firms can require approval in Claira above a threshold they set.
The session carries your own case access and field security, so the assistant can’t do anything you couldn’t do in Claira.
Tool calls are logged with the user, the assistant and the time, and reviews started from an assistant are marked in the task history.
Documents in a case are written by the other side, so their text comes back as clearly marked data, the assistant can’t invent a set of documents to act on, and nothing that uses tokens or writes runs without a fresh confirmation.
Join early access.
We’re opening Claira MCP to a small group of law firms, legal departments and service providers. Tell us which assistant your team uses and what you’d run through it first.
- English, French and German
- Your data residency, kept
- Off by default, controlled by your admin
- Uses your existing plan and tokens
Questions about Claira MCP
A Model Context Protocol (MCP) server exposes a product’s capabilities as tools an AI assistant can call. Claira MCP lets Claude, ChatGPT or any MCP client check bulk review status, launch privilege and relevance review, query results document by document, and pull cited insights from a case already connected to Claira.
By default the Claira MCP server returns results, statuses, citations and metadata, never document text. Claira’s own models, in your residency region, are the only AI that reads your documents. An administrator can choose to allow citation excerpts or full text for a case, and that choice is logged.
Claira MCP is a remote MCP server using Streamable HTTP and OAuth 2.1, the pattern Claude and ChatGPT connectors expect. Claude Desktop, claude.ai, Claude mobile and ChatGPT are the tested clients, and any other MCP-compliant assistant can connect the same way.
No. Calls that use tokens or write to a review field happen in two steps: the server returns a preview with the document count, the tokens and the destination fields, and nothing runs until you confirm. Firms can also require approval in Claira above a token threshold they set.
It uses the same plan and tokens as the Claira app, whether a review is started in Claira or from your assistant.
An API connects software to software and needs someone to build the integration. An MCP server connects your evidence to the assistant a reviewer already has open, in plain words, with confirmations and an audit trail built in.
Try it on your next case.
Sign up with your work email, or book 15 minutes with the team.