Workboard

Page scope: Workboard coordinates multiple agents on tasks. For governed skill files see Skill Workshop. For lightweight delegation see sub-agent setup.

Workboard landed in stable v2026.6.1 alongside Skill Workshop. It gives you a board-shaped view of multi-agent work: tasks spawn runs, agents hand off across scoped session namespaces, and stale lifecycle issues were tightened in v2026.6.5. Use it when one chat session cannot represent a whole project.

• Task-backed runs — Work items on a board drive agent execution, not ad-hoc DMs alone.
• Multi-agent handoffs — Different agents (or namespaces) own steps in a pipeline.
• Scoped namespaces — Agent/global session namespaces with exact namespace tool dispatch (v2026.6.1+).
• Code mode — Same release train added task-backed coding workflows on the board.

• You run agent squads—researcher, writer, reviewer—on shared goals.
• Sub-agents in one channel are not enough visibility; you want a board UI.
• You coordinate long-running projects with explicit task state.

Skip Workboard if a single agent + cron + multi-agent team patterns already cover your load.

1. Upgrade to v2026.6.1+ (prefer v2026.6.5 stable): openclaw update, openclaw doctor --fix.
2. Confirm multi-agent routing in configuration—separate agents need isolated workspaces where required.
3. Open Workboard in Control UI (label may vary by build).
4. Create a small test task (e.g. “summarize three URLs, then draft a post”).
5. Watch namespace handoffs; confirm each agent has the tools it needs (tools.profile / tools.allow).
6. After upgrade to v2026.6.5, re-test stale task cleanup if you leave boards open for days.

• Add Workboard to post-upgrade smoke tests: upgrade runbook → v2026.6.1+.
• Board runs multiply token use—assign cheaper models to mechanical steps (cost examples).
• Keep pairing tight if board-triggered agents can message external channels.