Home
Get a saved, cited report from your AI agent — not a chat message you lose.
Cool Workflow (CW) is an auditable workflow control-plane for AI coding agents. Point it at a repo or
a folder of docs and it plans the work, delegates each task to your agent, records and verifies every
result, and writes a report where every claim is cited to a file:line — all as durable .cw/ state
on your own disk.
CW never runs the model itself. The model is fuel; CW is the black-box recorder. Your agent spends the tokens; CW keeps the books.
| Page | Use it for |
|---|---|
| Getting Started | Install, run the 30-second tamper proof, and produce your first cited report. |
| Mental Model | Why CW is built the way it is — the four commitments, and when it's worth it. |
| Glossary | Every core term in one place: run, evidence, topology, verifier gate, and more. |
| Quickstart | The fast command reference once you know the shape. |
| Page | Use it for |
|---|---|
| Workflow Apps | Choose between the shipped apps and inspect app contracts. |
| Architecture | The runtime boundary, state files, verifier gate, and MCP surface. |
| Trust And Audit | What telemetry, audit verification, and the trust limits actually prove. |
| Recovery And Restore | Resume, export, inspect, import, verify, and rerun durable runs. |
| Commands or API | The stable CLI shapes and MCP entry points. |
| MCP And Manifests | Generated vendor manifests and CLI ↔ MCP parity. |
| Operations | Verify, restore, regenerate manifests, and run release checks. |
| FAQ | Trust limits, agent setup, reports, and failure behavior. |
- plans workflow apps into phases and tasks,
- dispatches isolated workers,
- delegates each worker to an external agent — it embeds no model SDK and holds no credentials,
- accepts and verifies
result.mdenvelopes, - records audit and telemetry ledgers (tamper-evident, ed25519),
- commits only verified state, and
- generates a cited report from the durable run record.
When evidence, dependencies, or rationale are missing, CW does not invent success — it stops at a
visible unexplained / blocked state. See the Mental Model for why.
This Wiki summarizes current repository evidence rather than introducing new public claims. The root README remains the official first-run page.