FAQ
Frequently asked questions
Short answers to the questions people ask first.
Is there a hosted version or an account to sign up for?
No. agent-ops is self-hosted and internal to the catesworks org. You run the control repo, the Lambda relay and the runner under your own accounts. There is no sign-in and no account.
Will it merge code without me?
Only dependency bumps and whitespace-only diffs, and only with a DONE completeness verdict and green CI. Everything else is held for human merge.
Does adding the agent-ready label start a run?
No. The label grants permission. A run starts only when a labeled task’s Status moves to Next.
Which coding agents does it support?
Claude (the default) and Codex. Copilot is declared but fails fast because it is not supported.
Is there an MCP server?
No. agent-ops is an MCP client: it calls the implement and review tools of mcp-agent-bridge, a separate local MCP server.
Does onboarding add workflows to my repo?
No. Onboarding adds four labels, a Project board link and, optionally, knowledge-base/config.yml. It never removes or overwrites existing automation.
Can I install the skill from npm?
Not today. The skill packages are private and the published-skills list is empty. The skill is distributed through the repository’s Claude Code plugin marketplace.
How many runs at once?
One self-hosted runner executes every dispatch and revision, so runs queue. The ledger records queue-wait, and ADR 0004 says when to add runners.
Still stuck? Contact us.