MCP · 01 / Your agents

Claude · Codex · internal agents · your runtime

Ask the question that belongs to the current workflow step.

MCP
What it is

The client side is whatever you already use. Claude Code and Codex are strong at reading a repository and editing it; connecting GitMir gives them somewhere to ask instead, when the question is about the product rather than one line. Cursor and VS Code read the same server from a config file, and anything that speaks MCP over HTTP can too.

What you do
01

Get your key from the Access page in the workspace.

02

Run the client's own add-server command, or paste the config block.

03

Restart the client and confirm seventeen gitmir_ tools are listed.

04

Ask the agent something about the product and watch which tool it reaches for.

What it does not do

Every one of these is "used side by side", not integrated: there is no GitMir plugin, extension or marketplace entry for any of them, and nothing is pushed into the client automatically. Codex cannot take an arbitrary header, so its key is read from an environment variable it re-reads on every start. Nothing here changes how your agent is orchestrated or permissioned.

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.