Use case
Runs on Local files
Hand issues to a fleet of agents
Label an issue, and the issue router packages it for a coding agent, opens a pull request with the result and moves the issue to review.
The problem
Running several coding agents at once turns into bookkeeping: which agent has which task, how many are running, which pull requests are waiting for review, and what happens when one fails.
How Kanbus helps
The issue router reads the board, picks eligible issues by label and class, and dispatches each one to a configured agent, such as Codex or OpenCode. It respects limits on work in progress per project, per class and per provider, watches the resulting pull requests and moves issues through your workflow's review and closed statuses.
kbs router plan
kbs router run --watch
kbs router status
You can pause the router, hold one class of work or one provider, and cancel a package without touching the rest.
What you set up
- Local files, with a
reviewstatus in the workflow. - A
router:block in.kanbus.ymlthat names the agent commands, the limits and the GitHub repository.