Skip to main content
The reference and concept pages tell you what each piece of Clawboo does in isolation. These guides do the opposite: each one is a single, realistic workflow that wires several features together: a team of mixed runtimes, a budget with a hard cap, an MCP attach to an outside agent, and walks it start to finish. Where a guide needs the precise shape of a route, store, or default, it links to the relevant concept, using, operating, or reference page rather than restating it. Start with whichever outcome matches what you are trying to ship; each guide names its prerequisites up front.

​
The guides

  • Build a multi-runtime team: connect more than one of the five runtimes (clawboo-native, openclaw, claude-code, codex, hermes), deploy a team that mixes them, and watch them coordinate over the board and group chat.
  • Set up governance & budgets: put a USD budget with a hard cap on an agent, mission, or team, layer on depth/fan-out/cost caps, and route risky delegations through the approval queue so a runaway run is stopped, not just logged.
  • Self-host securely: keep the default loopback bind, turn on the access gate, understand the encrypted credential vault and redaction, and expose the dashboard safely if you must reach it off-machine.
  • Attach the MCP to an external agent: point an outside agent (a Claude Code, Codex, or any MCP client) at Clawboo’s Tasks, Memory, Tools, and TeamChat servers so it joins your team as a peer over the shared plane.
  • Cross-runtime handoff: pause a task on one runtime and resume it on another, carrying the worktree system-of-record and AGENT_HANDOFF.json so the second runtime picks up cold.
  • Recurring team work: schedule team-task work with Routines, understand the team-task vs runtime-own-life cron domains, and handle the error-halts-until-resumed posture.

​
See also