Been experimenting with multi-agent workflows (OpenClaw on personal machine + Claude Code on work laptop) and kept hitting the same issue:
everything works⦠until you switch machines or try to plug systems together for managikng yoru tasks across systems. So I built a small system (Threadron) and wired it up via MCP.
What it does:
- shared work items (goal, current state, next step, blockers)
- append-only timeline (who did what)
- artifacts (PRs, plans, outputs)
- simple claim system so agents donβt step on each other
Now OpenClaw can:
- tell you about your task status across a multi agent multi system setup
- pick up work started by another agent
- see what already happened
- continue without re-deriving context
Setup is just MCP:
openclaw mcp set threadron '{"url":"https://threadron.com/mcp","headers":{"Authorization":"Bearer YOUR_API_KEY"}}'
Run it on your own at https://github.com/jerednel/threadron
Flashy site at https://threadron.com