Set up for team repositories

kipp•punkt is designed for a small number of high-performing people who cover the full value stream end-to-end. In team setups, optimize for clarity and control, not broad participation.

Assume a small owner group

The best setup is one or just a few core owners who drive the full loop: refining issues, approving requirements, and merging PRs. Collaborate when it makes sense (strategic discussions, discovery, a second opinion on a tricky PR), but the default mode is autonomous: one person and the agent, from idea to shipped code.

Control trigger permissions with `allowlist`

By default, everyone with repo access can trigger agent actions. Configure allowlist to restrict this to specific users:

{
  "allowlist": ["alice", "bob"]
}

This prevents drive-by comments from changing execution flow.

Use `requireMention` to separate discussion from action

In shared issues and PRs, people often discuss ideas without wanting agent action on every message. With requireMention set to true, the bot reacts only when explicitly @mentioned:

{
  "requireMention": true
}

This keeps normal human discussion clean and makes bot activation intentional. Especially useful when issues have multiple participants who aren’t all driving the agent.

One runtime per repo

Multiple runtimes cause collisions, duplicate reactions, and inconsistent state. If the agent needs to run on a shared server, coordinate who starts and stops it.

Scale throughput by tuning one runtime

If you need more speed, scale the single runtime instead of adding more instances. Tune maxConcurrency up and pollIntervalMinutes down, while staying within GitHub rate limits:

{
  "maxConcurrency": 3,
  "pollIntervalMinutes": 3
}

This gives you more parallel work without coordination chaos. See Set up your repository for more on tuning these fields together.