Coworker moves bulk reading and boilerplate writing to a cheaper model so the reasoning model keeps its context for reasoning. It is a tool, not a policy: it is used where delegation is genuinely cheaper than doing the work in place, and skipped where it is not.
- Status
- Live
- Autonomy (AAL)
- L1 → L2
- Interface
- Command line
- Repository
- Arcanada-one/coworker
01
Context kept for reasoning
Bulk reading and boilerplate writing go to a cheaper model, so the reasoning model spends its context on reasoning.
02
A tool, not a policy
It is used where delegation is genuinely cheaper than doing the work in place, and skipped where it is not.
Capabilities
Working now
- Delegates bulk file reading and summarisation to a secondary model
- Vendor-neutral — the backend model is configuration, not a dependency
- Command-line interface usable from agents and from a terminal
Current autonomy level
Weakest link
Delegation quality depends on the backing model; a task that needs judgement rather than volume is not a good fit and the tool does not detect that for you.
Roadmap to L2
- Broaden the set of backends that can serve as the bulk reader
- Record delegation outcomes so the cheaper/not-cheaper call stops being a guess
Next step
Explore the available product and its current limits
LIVE means the project is available now; the weakest link and roadmap above state the current version’s limits.
Review the path to the target level