Start here: how to contribute, and what we need

[moeen-mahmud]

Thanks for looking at Dispach. This is the place to start if you want to help.

What Dispach is. A self-hosted agent runtime your product calls over HTTP: memory, tools, schedules, channels and approvals behind one API, on any OpenAI-compatible model. The docs cover using it; Limits and evidence says what it does not do.

Where help is most useful right now

• good first issue: small, self-contained, each saying where to start and which test proves it.
• help wanted: larger pieces with a design note: a new channel, a model transport, a published plugin entry point, and the small-model and evaluation work.
• Running Dispach against a model or endpoint we have not tried, and telling us what broke.

The roadmap says what is planned, where Dispach is behind, and what is deliberately not planned.

How a change gets in

1. Set up a checkout: development setup.
2. Comment on the issue you are taking, so two people do not do the same one.
3. Open a pull request; the template lists the checks (bun run lint, bun run typecheck, bun test).
4. A new manifest field, event, route or dependency changes a published contract: open a Discussion in Ideas before building it.

CLAUDE.md in the repo root is the standing brief: the hard rules and the mistakes the code base has already paid for. It is written for coding agents and is the fastest way in for people too.

Questions go in Q&A. Security problems go through private reporting, never a public thread.

Everyone here follows the Code of Conduct.