Skip to content

Multi-repo coordination

You're working on feature X in project-a and need to know how project-b shapes its API. Without repowire, you cd over, read code, summarize, paste it back. With repowire, the agent in project-a asks the agent in project-b directly.

Setup

Open both projects in tmux windows:

# window 1
cd ~/projects/project-a && claude

# window 2
cd ~/projects/project-b && codex

Both auto-register on SessionStart. They land in the same circle (shared tmux session) and can address each other by display name.

The ask

In project-a, tell your agent:

Ask project-b which endpoints accept multipart uploads.

The agent calls ask("project-b", "..."). The question lands in project-b framed [ask #cid from @project-a] .... project-b reads its own code, answers, and calls ack(cid, "POST /uploads and POST /imports/csv, multipart/form-data, 50 MB cap").

Back in project-a, the reply arrives as a notification. The agent integrates it into the work in progress.

When this beats copy-paste

  • The other repo's code is the source of truth. You don't want a paraphrase you wrote five days ago.
  • You want a fresh read — local notes rot, the code doesn't.
  • The other agent has its own context (libraries, internal conventions) that would take you many turns to load.

When it doesn't

  • The answer is already in well-maintained shared docs. Reading docs is cheaper than asking.
  • The question is too vague to answer quickly. Asks should be specific enough that the other agent can ack in one turn.

Variations

  • Quick fact: bare ask, expect a one-line ack.
  • Pull a snippet: ask for the exact function signature plus where it's defined. The agent returns the path and line range; your agent reads it locally.
  • Coordinate a change: ask("project-b", "if I change the response shape of /users, what breaks on your side?"). The other agent runs a grep and answers.

See also