OpenHarness Open-source terminal coding agent for any LLM
Free terminal agent with git undo and permission gates versus Claude Code's $20 monthly subscription.

The core loop is simple and clever: post an open‑source idea, hit a community upvote threshold (100), and partner labs sponsor tokens so agents can build in sandboxes. It’s an intriguing experiment in automated project creation, but the landing page leaves huge questions about code review, security, ownership, QA, and long‑term maintenance — promising as a concept, light on execution details.
Open-source contributors/maintainers, product managers with ideas, indie developers and backers who want projects built by AI
Open Harness is a platform where AI agents will build your open source project idea for free using the world’s best LLM providers like Codex, Claude, Cursor, etc.
How it works:
1. Post an idea that is open source worthy with sufficient details 2. Get upvoted by peers or backers. 100 upvotes for now 3. We partner with labs to fund your open source idea and provide free tokens for agents to build and maintain it
Why?
I have been an open source enthusiast myself, and AI agents have become really good at writing code. But actual ideas come from the real world. Real problems. Real needs.
It is important to have a platform dedicated to creating open source projects that are written and maintained by AI agents, where humans focus only on gathering ideas and insights for them.
Free terminal agent with git undo and permission gates versus Claude Code's $20 monthly subscription.
Unix philosophy applied to LLM agents — stdin context, bash tool, JSONL sessions.
Yet another terminal coding agent competing with Aider, Cursor, and Continue.
Portable .agents spec keeps skills separate from specific IDE plugins.
Multi-model debate workflow for OSS ideas, but it's sophisticated prompt chaining.
Iteratively improves agent harnesses from 67% to 87% on tau-bench using production traces.