AgentMachinist
github.com/vscarpenter/agentmachinist- Category
- AI Agents
- Rank
- No. 1873Tools index
Previous survey · No. 1832 ·
- Pricing
- Open Source
- Type
- TOOL
- Use case
- Coding
- GitHub
- 7 stars
- Latest release
- v0.21.0
- Date
About
AgentMachinist takes a small development task from intent to a reviewed local change by coordinating coding agents such as Claude Code, OpenCode, Pi, Codex or Goose. A human approves the exact spec before implementation and reviews the result before integration. GitHub and GitLab are optional sources of tasks and destinations for publication.
At a glance
AgentMachinist gives coding-agent work an approval step before edits and a review step before integration. It is worth trying if you want to inspect the exact plan, diff and verification results before accepting a change into your repository.
Good fit
Developers who delegate bounded repository changes and want a repeatable approval workflow.
Consider something else when
Skip it if you want unattended merges or already have a lightweight review process that meets your needs.
Start here
Run machinist rehearse --guided in its disposable sample. Read the plan, approve it, then check the diff and review before choosing whether to integrate.
Cost and requirements
MIT-licensed. The CLI requires Python 3.12 or later and Git on macOS or Linux. The rehearsal uses a fake agent with no model calls. Real work needs a supported coding-agent CLI and may incur provider charges.
Limits to know
The separate review is part of the tool workflow. You still choose whether to integrate the reviewed candidate. Publishing to GitHub or GitLab is optional and requires its authenticated CLI.
Related work
Sources checked 2026-11-05. Source-based assessment.
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.