Skip to content

Slipway

Give local AI-agent work a durable lifecycle, current-state authority, and evidence-backed closeout — inside the repository where the work happens.

Slipway is a small governance control plane for local AI-assisted development. It does not replace your AI coding tool, project tracker, or Git — it makes agent work legible by binding every change to a lifecycle, a current authority file, and evidence that can be inspected after the session ends.

You drive it in plain language. After a one-time slipway init, you tell your AI agent what you want to build, and the generated adapter routes that request through the governed lifecycle — no command sequence to memorize. Slipway stays the authority on whether the change is actually done; the commands throughout these docs are what the agent runs for you, and what you can run directly whenever you want.

Watch demos

These short recordings show the two pacing modes most users care about first: hands-off progress when authorization is still fresh, and hard stops when operator judgment matters.

Auto mode. Pacing pauses advance when prior authorization remains fresh.
Interactive mode. Checkpoints stop where operator judgment matters.

Explanation

The design reasoning behind the governed workflow.

Design · Workflow

Completion is deliberately hard to fake: every governed stage owns evidence that the engine re-derives instead of trusting. Start with the tutorials to experience that loop, then reach for the reference when you need exact command syntax.