Autonomous Forge is a Python-based local-first CLI project for AI-assisted repository maintenance, framed as an experimental safety framework. It reports a large codebase evolved during an experiment, with test failures on main branch and no production readiness.
Collecting history — the radar snapshots this repo daily. The trend line appears after 3 days of data (1 so far).
What it is
Autonomous Forge is an open-source Python tool built and maintained by scheduled AI agents. The agents were given a GitHub repository and permission to decide what to build, how to structure it, and how to improve it.
This README documents what happened during the experiment.
How it works
The project provides local planning, safety checks, side-effect gates, and evidence records. It does not call an AI model directly; the external scheduler and AI agents provide autonomy. It supports planning, review and validation, controlled changes, and evidence/preservation workflows. The README includes pipelines for planning, proposal generation, validation previews, and guarded operations (apply patch, run validation, commit, push, archive).
Getting started
"Use it as": a case study of AI software stewardship, a reference for human-in-the-loop maintenance gates, a starting point for safer repository automation, or a testbed for evidence and replay design. Do not yet use it as an unattended tool for important repositories.
The README does not provide explicit installation or usage commands within the truncated content presented here.
Recent releases
Release entries: none.
Traction
Stars: 249 Forks: 9 Open issues: 4 Created: 2026-07-06 Last push: 2026-07-12
CI and test results indicate a mix of successes and failures across various Python versions, with the final local audit showing 569 passed, 82 failed, 1 skipped. The Actions history shows 242 successes, 1,153 failures, and 1 action required across runs.
Behind the repo
No startup or company information is provided beyond the repository itself.
Caveats
License: Apache-2.0 Language: Python Created: 2026-07-06 Last push: 2026-07-12 Tests show failing main-branch tests and no production-ready status.