Axocoatl CLI
A Rust host for coding agents: a dashboard, required checks, review by another model and a signed receipt for each turn
- 74 ★ on the repository
- Last commit
- Apache-2.0
Broad access
Why this level
- The agent writes and changes repository files, and runs unattended in headless mode
- Hosted providers such as OpenRouter receive the code context sent to them
- The installer sets up Podman and on macOS creates a virtual machine
Install
- Works in
Manual install
curl -fsSL https://axocoatl.ai/install.sh | shThen axocoatl onboard --start none and axocoatl doctor. The installer asks before installing Podman.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The agent writes and changes repository files, and runs unattended in headless mode
- Hosted providers such as OpenRouter receive the code context sent to them
- The installer sets up Podman and on macOS creates a virtual machine
This is third-party code. Review the repository files before installing.
What it does
Axocoatl is a single Rust executable that runs coding agents unattended and lets the host decide when the work is done. The axocoatl dev command starts a dashboard where one session holds the conversation, files, preview, terminal and Git. You open a repository, ask for one concrete change and read the real diff in the Files tab before committing. The host runs the configured checks on the final files, then a separate read-only activation performs the required review. The writer can be Ollama or OpenRouter in a dashboard session, plus Claude Code or Codex in headless mode: axocoatl run fix --task starts the built-in fix loadout, and fix-claude-code and fix-codex cover the external agents. Repository tools run in local rootless Podman by default. After a turn the dashboard offers a signed receipt to download. It records the verdict, models, checks, review and hashes of the covered files, and the signature can be verified offline with axocoatl receipt verify. The signature shows which key vouched for those facts and does not prove the change is correct. There is no Axocoatl account and no product telemetry.
Who it is for. Developers who hand small bug fixes to an agent and want to see checks, review by a second model and a record of how the work went.
Good fit when
- A pagination bug or failing tests in one module go to an agent, and you want review and check results before the commit
- A weak local model writes code and a stronger model checks it before acceptance
- You want to compare several solutions to one task and keep the best
- A team wants a receipt with file hashes and the review verdict that can be verified offline
Not a fit when
- You want the familiar Claude Code terminal interface: Axocoatl brings its own dashboard and sessions
- You need a guarantee the code is correct: the receipt confirms the facts of a turn and says nothing about the correctness of the change
- The platform is not a verified one: macOS on Apple silicon and Ubuntu 24.04 were verified, others were not
Example request
Run axocoatl run fix --task with the pagination bug description and verify the receipt with axocoatl receipt verify.Limitations
The verified platforms are macOS on Apple silicon and Ubuntu 24.04, and the author has not yet checked others end to end. The installer asks before installing Podman and creating a virtual machine on macOS. A signed receipt is not proof that the code is correct, it confirms recorded facts. If signing or storage fails, the command reports receipt_unavailable.
The author cites one small test on six tasks with local models: a required review by a stronger model raised the result from 17 to 19 of 25 withheld tests at 2.6 times the token cost. Review through OpenRouter and the built-in loadouts were not measured. Hosted providers receive the context sent to them.
How to disable. Stop the axocoatl dev process and delete the executable along with the user configuration and data directories. If you installed the daemon service, disable it first. Remove Podman and its virtual machine on macOS separately if you no longer need them.
FAQ
How do I install Axocoatl?
Run curl -fsSL https://axocoatl.ai/install.sh | sh, then axocoatl onboard --start none and axocoatl doctor. Setup covers the current OS user and creates no projects.
Which agents can Axocoatl run?
Ollama or OpenRouter write in a dashboard session, and in headless mode Claude Code and Codex are also available through the fix-claude-code and fix-codex loadouts.
What is a loadout in Axocoatl?
It is a recipe for a run: writer, checks, review and budgets. The built-in fix loadouts require the repository check and a review, while other loadouts have their own acceptance rules.
How do I verify a receipt?
Use axocoatl receipt verify with the path to receipt.json, the key from axocoatl receipt key, the --repo flag and optionally --require-pass. Verification works offline.
Does Axocoatl send data to its developers?
No, there is no account and no product telemetry. Hosted providers such as OpenRouter receive only the context you send them.
Discussion
No questions or reviews yet.
Related
CLI toolsOctomind is a coding agent in one Rust binary: safety rules in TOML instead of approval prompts, with a supervisor model on top
A lightweight Rust coding agent for the terminal, inspired by pi and OpenCode and built for a small memory footprint
Viden is a terminal cockpit for a coding agent: chat with a model, approve edits and supervise Codex and Claude in one window.
Локальная среда для команды агентов: проекты как каталоги, промпты как файлы и журнал каждого шага, который можно проверить.
A Rust CLI where a meta-agent hands work to Claude Code and Codex, each on its own branch and tmux session, in one view.
A terminal coding agent in one Rust binary that reads code, edits files and runs tests with any model of your choice.
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| Axocoatl (this card) | no | 74 | ||
| Ralph Orchestrator | no | not required | 3,203 | loops an agent until the task is done, while this one adds required checks, review by another model and a signed receipt |
| Orchestrator | no | 119 | distributes work across several agents, while this one builds the turn around host checks, a dashboard with a diff and a receipt |
README in short
The description starts with running agents unattended and walks from the first session to downloading a signed receipt. It then compares the working modes, covers the headless fix run and gives measured results with the author caveats.