Sidecat Work becomes a durable, governed record — its intent, target, and the conditions that must hold — instead of a message buried in a chat log.
Governed AI-assisted work
/\_/\
( o.o ) Sidecat
> queue -> run -> review A control plane for AI-assisted work you can trust.
Sidecat keeps serious AI-assisted work out of the chat log. Work is queued as a durable, governed record, run with explicit authority, and reviewed against evidence — so the assistant, the operator, and the next session all refer to the same work instead of rebuilding it from a transcript.
Sidecat is early R&D, built in the open. It is a working direction, not a packaged release or a hosted product.
The loop
Queue, run, review — over one durable record.
The assistant acts through explicit authority. External effects like commits, pushes, and deploys leave receipts, and what happened is kept next to what the agent was actually shown.
One service-owned record turns into review, replay, and a resume brief — so the work can be checked, corrected, and picked up later without transcript archaeology.
What it is for
Use Sidecat where work needs governance, not just execution.
When the work has to be trusted afterward
The hard part is not whether an agent can act — it is whether the change can be trusted, audited, and recovered if it went wrong.
When more than one mind shares the work
Operators and multiple agents need one durable place to refer to, so nobody rebuilds the state of the work from a transcript.
When authority and evidence matter
Commits, pushes, and deploys spend trust. Sidecat routes them through explicit authority so the receipts and evidence stay first-class.