teamclaw-setup（团队爪设置）

TeamClaw Setup

Guide users to the smallest working TeamClaw installation first, then expand.

Default workflow

1. Prefer the guided installer first:

   npx -y @teamclaws/teamclaw install

1. Use manual plugin install only when the user wants direct control:

   openclaw plugins install @teamclaws/teamclaw

If the user explicitly wants the ClawHub package path, use:

   openclaw plugins install clawhub:@teamclaws/teamclaw

1. Recommend controller + localRoles for the first successful run unless the user explicitly needs distributed or on-demand workers immediately.

1. Read references/install-modes.md before generating config. Pick the smallest matching topology and reuse the provided snippets.

1. Read references/validation-checklist.md before finishing. Always give the user:

- one health-check command - one UI URL - one tiny smoke-test requirement - one or two likely failure checks for the chosen topology

Installation guidance

• Start with a single machine and localRoles.
• Keep TeamClaw taskTimeoutMs large for real model runs.
• Keep OpenClaw agents.defaults.timeoutSeconds at least as large as the TeamClaw task window in seconds.
• For docker or kubernetes, require a reachable workerProvisioningControllerUrl.
• For docker, mention the published runtime image:

  ghcr.io/topcheer/teamclaw-openclaw:latest

What to produce

When helping a user, produce:

1. the exact install command
2. the minimal config snippet for the chosen topology
3. the startup command
4. the validation commands and first smoke-test task

Do not push users into distributed, Docker, or Kubernetes first unless they asked for it or already have the surrounding infrastructure ready.