Axocoatl — AI skill for Claude Code
Run coding agents unattended.
How to install Axocoatl
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open axocoatl/axocoatl and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Axocoatl does
Run coding agents unattended. Agents write the code. Axocoatl writes the receipt. Claude Code, Codex or a local model in a sandbox; the host runs your checks, a different model reviews, and every run ends with a CI exit code and a signed receipt.
Alternatives in AI
- Rb Setup — First-time setup 1.3k ★
- Pi Agent VS OpenCode — Customization & Control Comparison — Pi v0.52+ vs OpenCode v1.1+ (Feb 2026) Thesis: Pi and OpenCode are both MIT-licensed, open-source, model-agnos 526 ★
- Agenvoy — Self-hosted AI agent harness in a single Go binary — writes, sandbox-tests and repairs its own tools, and lets 474 ★
README
Axocoatl
**Run coding agents unattended, and let the host decide when they are done.** Agents write the code. Axocoatl writes the receipt.
Choose Claude Code, Codex, a local model or an OpenRouter model to make a change, and a different model to review it. The built-in fix loadouts run the writer in a sandbox, keep its model credential on the host, and have the host run your checks on the final files. An exit code reports the outcome; a signed receipt lets someone who did not watch verify the recorded facts and covered files offline. One open-source Rust executable; no Axocoatl account or telemetry.
[](https://github.com/axocoatl/axocoatl/actions/workflows/ci.yml) [](https://crates.io/crates/axocoatl-cli) [](LICENSE)
In the first terminal:
curl -fsSL https://axocoatl.ai/install.sh | sh
axocoatl onboard --start none # configure the writer, reviewer and sandbox
axocoatl doctor
axocoatl dev # leave this terminal running
In a second terminal, from the repository you want to change:
cd ~/src/my-app
AXO_EVIDENCE=$(mktemp -d "${TMPDIR:-/tmp}/axocoatl-first-fix.XXXXXX")
# Local or OpenRouter writer; choose the corresponding external loadout otherwise.
axocoatl run fix --task "Paging skips the last item when the total is a multiple of the page size" \
--junit "$AXO_EVIDENCE/junit.xml" --record "$AXO_EVIDENCE/run.axorecord.jsonl" \
--receipt "$AXO_EVIDENCE/receipt.json"
axocoatl receipt verify "$AXO_EVIDENCE/receipt.json" \
--key "$(axocoatl receipt key)" --repo . \
--record "$AXO_EVIDENCE/run.axorecord.jsonl" --require-pass
Follow [Your first verified fix](https://docs.axocoatl.ai/start/first-fix/) for the Claude Code, Codex, local and OpenRouter setup choices and expected result. Evidence stays outside the re
Related Skills
Oh My Dag
The orchestration layer under your coding agent. Turns work into a typed graph, runs one model per node, and t
Agents Sandbox
Run AI agents(Claude Code, Codex, OpenClaw, etc.) with Full Power in Isolated Local Sandbox. Zero cost, Zero r
Agentic Sandbox
Run AI agents(Claude Code, Codex, OpenClaw, etc.) with Full Power in Isolated Local Sandbox. Zero cost, Zero r
HaiKrew
Token-efficient agent crews for Claude Code: Haiku writes, the main model specs and reviews, a live pixel crew
Enclave
Sandbox for running AI coding agents autonomously: isolated, network-restricted, host-safe
Orbstack Sandbox
Sandboxed OrbStack VMs for LLM coding agents — locks down host filesystem access so agents in permissive modes
Related Agents
External Reviewer
Gets an independent second opinion on a finished resume or application from a different model (OpenAI Codex CL
Paf Reviewer
Runs one phone-a-friend relay or review command in the background and reports back a receipt plus the findings
Codex Runner
codex CLI(codex exec / codex exec resume)を最後まで走り切らせる専任エージェント。呼び出し元(メイン Claude / スキル)からプロンプト・model・effort・sandb