Archon Workflow Factory

What does this agent do, and when should you use it?

Archon is a self-hosted workflow engine where YAML describes processes built from agent steps, scripts, commands, loops, child workflows, and human decisions. It runs agent steps through the CLIs or SDKs for Claude Code, Codex, and Pi, and can give each repository run its own Git worktree. Its bundled sdlc pack moves GitHub issues toward reviewed, CI-checked pull requests that are ready to merge; runs can pause for answers in an attention queue. Users start workflows from the CLI and can inspect runs or answer questions in a web console; SQLite is the default database, with PostgreSQL also supported. It fits software teams that want self-hosted, repeatable procedures and approval gates; the bundled GitHub flow requires GitHub CLI.

Reads YAML workflows from .archon/workflows/ and executes nodes in dependency order, including prompts, bash commands, scripts, loops, and child workflows. Agent steps can run Claude Code, Codex, or Pi; loops continue until a specified bash condition succeeds or an iteration limit is reached. When a human decision is needed, a run pauses and creates an attention item with context, a recommendation, and choices, then resumes after an answer. The bundled sdlc pack uses gh to inspect issues, write a delivery brief, plan and implement changes, review and fix findings, wait for CI, then let a merge queue check the review verdict and CI before handling a pull request. Archon records events, transcripts, and run costs; repository runs use separate Git worktrees by default.

1. A maintainer who wants a fixed path from issue to pull request can use the bundled sdlc workflows for planning, implementation, review, and CI.
2. A software team handling several repository tasks at once can isolate each run in its own Git worktree.
3. A maintainer who wants queues to move work forward and request input only when needed can use the attention queue for paused runs.
4. A team with repeatable content, research, operations, or support procedures can encode their steps, checks, and human decisions as YAML workflows.
5. A team that wants run state and audit records on its own machine or server can self-host Archon with SQLite or PostgreSQL.

How do you install or deploy this agent?

Install Claude Code and sign in, or install Codex; the Archon installer asks which providers to add. On macOS or Linux:

curl -fsSL https://archon.diy/install | bash

On Windows PowerShell:

irm https://archon.diy/install.ps1 | iex

With Homebrew:

brew install coleam00/archon/archon
archon setup

The README also documents source, Docker, and server installation, but does not include their commands here. The bundled sdlc pack additionally uses GitHub CLI gh; triggers are not bound by default.

How do you use this agent?

From a Git repository, list workflows and start an investigation:

cd your-project
archon workflow list
archon workflow run archon-investigate "Why does the login test fail?"

You can also add a YAML file under .archon/workflows/, for example:

name: build-feature
description: Plan and implement a feature.
nodes:
  - id: plan
    prompt: "Read the request and code. Write an implementation plan."
  - id: implement
    depends_on: [plan]
    loop:
      prompt: "Implement the next step of the plan."
      until_bash: "bun run test"
      max_iterations: 10
      fresh_context: true

Start it with archon workflow run build-feature "Add dark mode". Run archon serve for the web console, or archon workflow attention to inspect and answer paused runs in the terminal. To let Claude Code or Codex drive Archon, run archon skill add in the project.

What are this agent's strengths and limitations?

How does this agent compare with similar options?

Key facts side by side with the most closely related agents.

How does FollowAgents rate this agent?

FAQ