Autonomous Forge is an open-source Python tool built by AI agents as a three-day experiment in safe repository maintenance, resulting in a pre-alpha human-in-the-loop safety framework with many features but a failing main branch.
251
No data
10
3
Apache-2.0
2026-07-12
This experiment showcases the ability of AI agents to autonomously grow a repository from two files into a structured project with extensive tests, safety boundaries, and engineering memory, while also revealing critical weaknesses in maintaining test stability and production readiness.
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.
The experiment was successful as a research test, but the project is not production-ready.
The AI grew a small repository into a large safety and maintenance framework. The code installs, compiles, and many individual commands work. However, the final main branch has failing tests. The latest GitHub Actions run failed on Python 3.10, 3.11, and 3.12.
The best description is:
A useful pre-alpha, human-in-the-loop safety framework — not a self-running AI engineer.
Autonomous Forge is a local-first command-line tool. It helps a maintainer or AI-assisted workflow move through a controlled maintenance process:
flowchart LR
A[Repository files] --> B[Plan and select a task]
B --> C[Create a reviewable proposal]
C --> D[Review paths, diffs, status, and risks]
D --> E[Preview validation]
E --> F[Optional confirmed change]
F --> G[Commit and push checks]
G --> H[Evidence bundle and replay]
H --> I[Archive and preservation checks]
D --> J[Human review]
J --> F
The repository does not call an AI model. The external scheduler and AI agents supplied the autonomy. Forge supplies local planning, safety checks, side-effect gates, and evidence records.
The repository started with only a README and a license. By the final commit, it contained:
.ai planning and memory files.AUTO-### task groups, reaching AUTO-140.| Local date | Commits |
|---|---|
| 7 July 2026 | 194 |
| 8 July 2026 | 658 |
| 9 July 2026 | 572 |
| 10 July 2026 | 61 |
xychart-beta
title "Commits by local date"
x-axis ["Jul 7", "Jul 8", "Jul 9", "Jul 10"]
y-axis "commits" 0 --> 700
This repo is automated and is fully AI-built and AI-maintained project that continuously plans, codes, tests, and improves itself.
Similar projects matched by category, topics, and programming language.
Harness Engineering is a methodology for improving coding agent outputs by carefully crafting the environment around them—providing curated context, tools, and executable constraints that encode an organization’s nonfunctional requirements and cumulative lessons.
A 28.9 million parameter language model runs on an $8 ESP32-S3 microcontroller entirely on-device, generating simple stories at about 9.5 tokens per second.
mimic captures traffic from any iOS or web app and automatically generates a Python client library that lets you call the app's API like a regular library.