Claude, Codex & coding agents
Retrieve task-specific product Intelligence instead of reconstructing the product from source on each prompt.
GitMir runs an MCP server, so any MCP-capable client connects to it — Claude Code, Codex, Cursor, VS Code with Copilot. The account's Access page has the exact line for each. Once connected, the client gets seventeen tools: ask what a part does, find something by name, see what a change reaches, read what changed between two builds, see what is worth a person's attention, and pick up accepted work. Those tools are good at writing code; this gives them somewhere to look the product up instead of inferring it again on every prompt.
Open Access in your account and copy the line for your client
Paste it into Claude Code, Codex, Cursor or VS Code
Restart the client and check it lists seventeen gitmir_ tools
Give the agent a task and let it call the tools it needs
Have it close the task with what the product now does differently
There is no plugin or extension to install — you paste one line of configuration, and each client has its own quirk the Access page walks through (Codex takes the key from an environment variable rather than a header; Cursor and VS Code want it in a JSON file). The tools only answer about repositories GitMir has already read, and there is no cross-repository question. Agent calls over MCP are not metered, but the rebuilds their commits eventually trigger draw on the same credit pool.
See it on a real product.
Three models are published in full and open without an account. Ask one of them the question this page is about.