ezyang/codemcp

codemcp fully obsolete. However, there are some good design ideas (especially around
the Git-versioning scheme) that I eventually want to port into the current generation
of agentic coding clis.
codemcp
Make Claude Desktop a pair programming assistant by installing codemcp. With
it, you can directly ask Claude to implement features, fix bugs and do
refactors on a codebase on your computer; Claude will directly edit files and
run tests. Say goodbye to copying code in and out of Claude's chat window!
codemcp offers similar functionality to other AI coding software (Claude Code,
Cursor, Cline, Aider), but it occupies a unique point in the design space:
1. It's intended to be used with **Claude Pro**, Anthropic's $20/mo
subscription offering. I like paying for my usage with a subscription plan
because it means **zero marginal cost** for agent actions; no more feeling
bad that you wasted five bucks on a changeset that doesn't work.
Note that if you have Claude Max ($100/mo), Claude Code can also be used
with subscription based pricing. The value proposition for codemcp is
murkier in this case (and it is definitely inferior to Claude Code in some
respects), but you can still use codemcp with Claude Max if you prefer some
of the other UI decisions it makes. (Also, it's open source, so you can
change it if you don't like it, unlike Claude Code!)
2. It's built around **auto-accept by default**. I want my agent to get as
far as it can without my supervision, so I can review everything in one go at
the end. There are two key things that codemcp does differently than most
coding agents: we **forbid unrestricted shell**, instead requiring you to
predeclare commands the agent can use in codemcp.toml, and we **Git
version all LLM edits**, so you can roll back agent changes on a
fine-grained basis and don't have to worry about forgetting to commit
changes.
3. It's **IDE agnostic**: you ask Claude to make changes, it makes them, and
then you can use your favorite IDE setup to review the changes and make
further edits. I use vim as my daily driver editor, and coding environments
that require VSCode or a specific editor are a turn off for me.
IMPORTANT: For master users - Major changes for token efficiency
To improve codemcp's token efficiency, on master I am in the process of
changing codemcp back into a multi-tool tool (instead of a single tool whose
instructions are blatted into chat when you InitProject). This means you have
to manually approve tool use. Because tool use approval is persistent across
multiple chats, I think this is a reasonable tradeoff to make, but if you
really don't like, file a bug at
refined-claude browser
extension for supporting auto-approve tool use.
Installation
I recommend this specific way of installing and using codemcp:
1. Install uv and install git, if they are not installed already.
2. Install claude-mcp on your browser.
This enables you to connect to SSE MCP servers directly from the website,
which means you don't need to use Claude Desktop and can easily have
multiple chat windows going in parallel. We expect this extension should
be soon obsoleted by the rollout of
Integrations. At time of
writing, however, Integrations have not yet arrived for Claude Pro subscribers.
3. Run codemcp using uvx --from git+https://github.com/ezyang/codemcp@prod codemcp serve.
You can add --port 1234 if you need it to listen on a non-standard port.