Archon is an open-source harness builder for AI coding: instead of prompting a coding agent ad hoc and hoping for a consistent result, you describe your development process as YAML workflows, and Archon runs them deterministically with agents, scripts and human approval gates in the right places. Its bundled software development lifecycle pack takes a repository from an open issue to a merged pull request. Archon has gathered roughly 23,700 stars on GitHub since its 2025 release.
Asking a coding agent to fix a bug gives a different process every run: steps get skipped, tests forgotten, templates ignored. Archon fixes the process and lets the agent supply judgment inside each step. Workflows are graphs of nodes: agent prompts, bash commands, loops that repeat until a test command passes, and waits where a run pauses and presents one clear attention item, with a recommendation, for a human decision. Built-in queues claim issues, run the plan-implement-review-fix pipeline, and land approved pull requests, with a full event log, transcripts and per-run cost tracking stored in your own database. Archon orchestrates existing agents such as Claude Code, Codex and Pi through their own CLIs and accounts rather than replacing them, is fully self-hosted, and can be driven from the CLI, a web console, Slack, Telegram or Discord. The project even builds itself: the team reports most of its merged pull requests are produced by Archon runs.
Archon is free and open source under the MIT license. You install it on your own machine or server; the only costs are the coding-agent subscriptions you connect to it.
Archon is for developers and small teams who already use AI coding agents and want repeatable, governed automation around them, turning one-off prompting into a reliable software factory they control.