Everything the work needs. Nothing off the record.
The questions a firm asks first, answered by the system itself.
Who decided?
A lawyer did. The AI makes the first pass — review, markup, extraction — and every suggestion waits for a named person's accept or reject. Nothing applies itself.
Who changed this clause?
The record says. Person or AI, each edit is saved with an author, a reason, and the exact before and after. Any clause's history is one click away.
Where did the documents go?
Through your own LLM account and back — Claude, Gemini, OpenAI, or OpenRouter, stored encrypted, set for zero data retention. Open source, self-hostable if a client requires it.
Nine tools. One history underneath.
Each tool does one job well. Research tools keep their queries, sources, and activity visible in the assistant session; the tools that change the matter write to the same record, so the work reads as one story no matter who — or what — did it.
Research the law and Australian IP without leaving the matter.
Search US case law and Australian patents and trade marks in the assistant or through your connected MCP agent. In Australia, queries, IP Australia sources, and tool activity stay visible in the assistant session.
Every document for the matter, in one working space.
Upload and organize agreements, correspondence, and supporting files. Read them, compare versions, share access with the matter team, and use the same source material across chat, review, redlining, and drafting.
Ask about the matter. Get answers with sources.
Work with the documents filed in a matter. Open a document or ask the assistant to examine the relevant files, with answers linked back to their sources. Runs on your firm's LLM key.
A hundred contracts. One pass.
Ask the same questions of every document in a stack and get the answers back as a table. Each cell links to the spot it came from and to who ran it. The outliers surface themselves.
The first pass is done before you open the file.
Mark up agreements against your own playbook. Suggestions land as tracked changes you accept, reject, or trace back — and drafts start from your templates, not a blank page.
Turn good process into a button.
Capture review, extraction, and drafting as simple, repeatable steps. Run them on the next stack of documents, share them across the team, and stop re-explaining the same job.
Your standard positions, in one place.
Approved clauses, fallbacks, and the playbooks the AI reviews against — shared across the firm, so every markup starts from the positions you've already decided on.
Every change is on the record.
Person or AI, every edit is a commit with an author, a message, and the exact before and after. Open any clause and see how it got there. Read it, share it, or undo it.
ChatGPT and Claude plug straight in.
Connect the assistant your firm already pays for over MCP — no new chatbot, no second login. The agent drives gitmatter's tools; gitmatter records every step it takes.
The AI your team is already using.
Connect as an agent
Over MCP, with a secure sign-in — the agent drives gitmatter's tools and every action lands on the record.
- ChatGPT
- Claude Desktop
- Claude web
- Claude Code
- Codex CLI
Run on your own key
gitmatter's own features run on the firm's LLM account — stored encrypted, configured for zero data retention.
- Anthropic Claude
- Google Gemini
- OpenAI
- OpenRouter
See the work. Skip the black box.
Book a walkthrough with your own AI account and see the record build itself: every change saved with who, what, and when. Set up in minutes.