Archon

Open-source workflow engine for AI coding agents by Cole Medin. Define development processes as YAML DAG workflows — planning, implementation, validation, code review, PR creation — and run them deterministically across all projects. Framing: “Like what Dockerfiles did for infrastructure and GitHub Actions did for CI/CD — Archon does for AI coding workflows.”

GitHub: coleam00/Archon
Website: archon.diy
License: MIT
Prerequisites: Bun, Claude Code, GitHub CLI

What Problem It Solves

When you ask an AI agent to “fix this bug”, what happens depends on the model’s mood. It might skip planning, forget to run tests, or write a PR description that ignores your template. Every run is different. Archon fixes this: encode your development process as a workflow. The workflow defines the phases, validation gates, and artifacts. The AI fills in the intelligence at each step, but the structure is deterministic and owned by you.

Core Concepts

YAML DAG Workflows

Workflows live in .archon/workflows/ and define a directed acyclic graph of nodes:

AI nodes — prompt-driven; the model does planning, code generation, or review. Optional fresh_context: true resets the context window each iteration.
Deterministic nodes — bash scripts, test runners, git operations. No AI involved.
Loop nodes — iterate until a condition is met (e.g., ALL_TASKS_COMPLETE). Can be AI loops (iterate on implementation) or human approval gates (interactive: true pauses for input).
Dependency edges — depends_on creates the DAG; independent nodes run in parallel.

Workflows are committed to the repo. Same-named files in the project override bundled defaults. The whole team runs the same process.

Git Worktree Isolation

Every workflow run gets its own git worktree on its own branch. Run 5 fixes in parallel with no conflicts. Empty worktrees auto-clean on close.

17 Default Workflows

Workflow	What it does
`archon-assist`	General Q&A, debugging, exploration
`archon-fix-github-issue`	Classify → investigate → plan → implement → validate → PR → review → self-fix
`archon-idea-to-pr`	Feature idea → plan → implement → validate → PR → 5 parallel reviews → self-fix
`archon-plan-to-pr`	Execute existing plan → implement → validate → PR → review → self-fix
`archon-ralph-dag`	The Ralph loop as a first-class workflow — iterate through PRD stories until done
`archon-piv-loop`	Guided Plan-Implement-Validate loop with human review between iterations
`archon-smart-pr-review`	Classify PR complexity → run targeted review agents → synthesize
`archon-comprehensive-pr-review`	5 parallel reviewers with automatic fixes
`archon-remotion-generate`	Generate/modify Remotion video compositions
`archon-architect`	Architectural sweep, complexity reduction, codebase health
`archon-refactor-safely`	Safe refactoring with type-check hooks and behavior verification
`archon-resolve-conflicts`	Detect merge conflicts → analyze → resolve → validate → commit

Run archon workflow list or describe what you want — the router picks the right workflow.

Multi-Platform

Platform	Setup time
Web UI + CLI	Out of the box
Telegram	5 min
Discord	5 min
Slack	15 min
GitHub Webhooks	15 min

The Web UI sidebar aggregates conversations from all platforms — workflows kicked off from CLI, Slack messages, GitHub issue interactions all appear in one monitoring hub.

Web UI

Chat — conversation interface with real-time streaming and tool call visualization
Dashboard — Mission Control for monitoring running workflows; filterable by project, status, date
Workflow Builder — visual drag-and-drop editor for creating DAG workflows with loop nodes
Workflow Execution — step-by-step progress view for any running/completed workflow

Architecture

Platform Adapters (Web UI, CLI, Telegram, Slack, Discord, GitHub)
                    │
                    ▼
              Orchestrator (Message Routing + Context Management)
               /         |          \
         Command     Workflow      AI Assistant
         Handler     Executor      Clients
         (Slash)     (YAML)        (Claude / Codex)
               \         |          /
                    │
                    ▼
         SQLite / PostgreSQL (7 Tables)
  Codebases · Conversations · Sessions · Workflow Runs
  Isolation Environments · Messages · Workflow Events

Lineage

Archon has pivoted twice:

v1 — AI agent builder (Python, RAG) — Cole’s original “all-in-one AI tool set for building agents.” Preserved on the archive/v1-task-management-rag branch.
v2 — AI OS for AI coding — knowledge + task management backbone via MCP (per WorldofAI walkthrough). Docker + Supabase + MCP server. Repositioned from the agent builder.
v3 (current) — Workflow engine — YAML DAG workflows, git worktree isolation, multi-platform adapters, Web UI. Bun-based. The knowledge/task backbone has been replaced by the workflow executor as the core primitive.

How It Compares

	Archon (current)	TASK.md pattern	bmad-method	Claude Code Ultra Plan
Form factor	YAML workflow engine + Web UI	Markdown files in project	Six markdown artifacts	CLI + cloud planning
Determinism	Full — YAML defines every phase	Partial — discipline is in the rules	Partial — discipline is in the personas	None — model decides
Isolation	Git worktree per run	Manual branches	Manual branches	Manual
Multi-platform	CLI, Web, Slack, Telegram, Discord, GitHub	Per-IDE	Per-IDE	Claude Code only
Parallelism	Built-in (independent nodes run in parallel)	Manual terminals	Manual	Sub-agents

AI For Dev

Explorer

archon-os