Canon

TIP

This wiki is aligned with Canon 0.63.0. For older versions, refer to the repository tags.

The governance runtime for AI-assisted engineering. Keep AI agents bounded, inspectable, and safely restricted to approved work zones.

Why Canon?

No Opaque Loops: You control exactly when agents plan, run, and publish.
Bounded Execution: Agents operate strictly within approved risk and zone limits.
Inspectable State: Every decision, approval, and output is captured as durable evidence.
Governed Packets: Turn unstructured chat into canonical, versioned markdown artifacts.

Canon operates on a simple, predictable four-step mental model:

init → Prepare the workspace.
run → Start a governed session with explicit boundaries.
approve → Review and unblock the agent when human judgment is needed.
publish → Promote the final artifacts into your repository's permanent memory.

Get your first governed session running in seconds:

bash

brew tap apply-the/canon && brew install canon
cd my-project
canon init
canon run --mode requirements --risk bounded-impact

These are the commands you'll actually use every day:

Command	What it does
`canon run`	Start a new governed session with explicit boundaries.
`canon status`	See exactly what the agent is doing right now.
`canon inspect`	Review generated evidence and artifacts.
`canon approve`	Unblock a session that hit a governance gate.
`canon publish`	Commit the final work into your repository.

Explore the wiki sidebar for advanced integrations, semantics, and architecture: