Central inbox
Collect pending AI review items in one working queue.
Review inbox
Separate a useful proposal from a trusted change.
TOW makes AI-generated proposals visible with context so people can inspect what would change before accepting it.
How it works
A clear path from suggestion to decision.
The inbox keeps review work in the system where the resulting change will live.
- 01
Propose
An enabled agent or AI workflow drafts a scoped change.
- 02
Inspect
A person reviews the target, rationale, context, and proposed diff.
- 03
Decide
Accept or reject the proposal and keep the decision visible.
Capabilities
Review without losing the operating context.
Keep proposal status and workspace state distinct.
Diff context
Show proposed changes against their intended workspace target.
Decision actions
Accept or reject with an explicit human action.
History
Retain proposal and decision state for later inspection.
Control point
Human review is the default boundary for proposals.
Reviewability does not guarantee a correct decision, but it gives teams a concrete control point between generation and accepted workspace state. Administrators can explicitly opt narrow agent workflows into direct writes.
- Visible proposal
- Named target
- Explicit decision
- Recorded outcome
Deployment and security
Control the deployment and the boundary.
Use TOW Cloud, run TOW on controlled infrastructure, or operate an air-gapped Docker Compose deployment. Permissions, identity settings, exports, and AI provider choices stay visible to administrators.
Explore self-hosting
FAQ
Questions worth answering clearly.
Current product, deployment, and commercial boundaries.
Does every AI answer require approval?
No. Informational answers can be read directly. Workflows that propose consequential workspace changes can use review-oriented inbox flows.
Can a reviewer reject a proposal?
Yes. The review flow distinguishes acceptance and rejection rather than treating generated output as automatically trusted.
See TOW in your environment