AI For Dev

archon-os

3 min read

Archon OS

Open-source command center for AI coding assistants by Cole Medin and the Dynamis community. Repositioned from Cole’s earlier “Archon AI agent builder” into the first-ever AI Operating System for AI coding — a knowledge + task management backbone any AI coding agent (Claude Code, Cursor, Kira, etc.) can plug into via MCP. Sleek modern UI for managing knowledge, context, MCPs, and tasks across projects, with a real-time visualization of agent task lists you can edit while the agent is working.

  • GitHub: coleam00/Archon
  • License: Open source
  • Stack: Docker + Supabase + your choice of model provider (OpenAI / Anthropic / Gemini / Ollama / OpenRouter)

What It Replaces / Augments

  • MCP context layer: instead of every agent having its own context loading, Archon exposes a unified knowledge base over MCP that all your agents can query
  • Task management: live task list per agent/project, editable in real-time — agent picks up your edits on the next iteration
  • Documentation indexing: continuously crawls and indexes external docs (e.g., Next.js docs, Shadcn docs) so agents always have current references — Cole frames this as “similar to Context7 but better”
  • Project manager dashboard: backlog → in progress → review → completed view per agent

Architecture (microservices)

  • Server — core API and orchestration logic
  • MCP server — the surface coding agents connect to
  • Agents service — agent execution (the wiki page calls this “coming soon” in the WorldofAI walkthrough)
  • UI — web dashboard
  • All ports configurable; everything runs via docker compose

Setup

  1. Prerequisites: Docker Desktop, Supabase project, git, VS Code, model provider API key (or Ollama for local)
  2. Clone coleam00/Archon, cd archon
  3. Copy .env.example → .env, fill in Supabase URL + credentials + model provider
  4. In Supabase SQL editor, run the contents of migration/complete_setup.sql
  5. Start Docker Desktop, then docker compose up
  6. Open localhost, configure provider + chat model + embedding model in Settings
  7. Optional: enable contextual embeddings, projects/tasks, dark mode
  8. Connect coding agents via the MCP Server tab — copy the MCP config snippet into your agent’s JSON config

Knowledge Base Workflow

The knowledge base accepts:

  • URL crawls — recursive ingest of documentation sites (the Next.js docs example from the video)
  • Codebases — point at a local repo
  • Business projects — structured project context
  • Plain URLs — single-page references

Once added, every connected agent can query the knowledge base in real time via MCP. This is the architectural answer to “how do I keep my agents on the latest framework docs without manually feeding them context every session.”

How It Compares

Archon OSTASK.md patternbmad-methodClaude Code Ultra Plan
Form factorHosted dashboard + MCP serverMarkdown files in projectSix markdown artifacts in AAI/ folderCLI + cloud planning
Knowledge layerContinuous crawl + indexingManual referencesManual referencesProject files only
Task UIWeb dashboard, real-time editableMarkdown checkboxesMarkdown storiesDoc-style web review
Multi-agent surfaceYes (any MCP client)Per-IDEPer-IDEClaude Code only
Setup costDocker + Supabase + configTwo markdown filesSix persona promptsClick /ultra-plan

Archon is the heaviest of the four but the only one designed to be a persistent, multi-agent knowledge backbone rather than a per-session pattern.

Lineage

The earlier Archon was Cole’s “all-in-one AI tool set for building agents” — referenced in his 100x faster video as one of the MCP servers he configures in Windsurf alongside filesystem, Brave, and git. Archon OS is the rebrand + repositioning into the AI-coding-OS frame.

See Also

Graph View

Backlinks