Staging draft, not for distribution
How it works

Repo in your copilot. Composed code out.

MCP is at once a local client and a hosted server, so two things are true at the same time: your code stays on your machine and you use a hosted compiler. Only bounded context crosses the wire, never the repository.

Your repo stays in your copilot

Your assistant (Claude Code, Copilot, Codex, Cursor) holds the repository and decides what a request needs. Pauhu never copies your repo.

Only bounded context crosses

Your copilot sends Pauhu just the bounded context the request needs. Not the whole repo, not your secrets.

The hosted compiler composes, zero tokens

Pauhu composes the code from cited operations and published standards. No model call, no GPU inference on the composed answer.

Composed code back, with a receipt

You get composed code in your copilot, and a receipt logs the tokens and energy the call avoided. Your model keeps only intent and composition.

Composed, not guessed

A compiler, not a model.

Your copilot was trained on the whole inconsistent public corpus, so it predicts code token by token and sometimes guesses wrong. Pauhu hands it the tool it lacked: the same request returns the same cited code every time.

The library of cited operations keeps expanding. Anything it has not composed, your own assistant handles as it does today, and the receipt shows which calls Pauhu composed at zero tokens.

1Your copilot, holding the repo
2Bounded context, not the repo
3mcp.pauhu.dev composes from cited ops
4Composed code, 0 model tokens
5Receipt: tokens and energy saved
Start free trial, 279 EUR/mo