Archon: The Operating System for AI Coding Assistants

AI coding assistants just levelled up. Archon turns scattered docs, brittle prompts and ad‑hoc to‑dos into a unified operating system for intelligent development.

It centralises knowledge, streams real-time context and drives task flow directly into Claude Code, Cursor, Windsurf and more, so code quality climbs and cycle time collapses. Setup is simple. Impact is immediate.

From messy debug sessions to AI‑powered clarity, Archon accelerates every iteration and amplifies every assistant.

Context

• Traditional AI Coding Setup: Lost between chats and tabs
• Archon OS: Centralized knowledge base with crawled docs, uploads and semantic search

AI Integration

• Traditional AI Coding Setup: One-off prompts per tool
• Archon OS: Model Context Protocol (MCP) server connecting Claude Code, Cursor, Windsurf, etc.

Knowledge Retrieval

• Traditional AI Coding Setup: Keywords and copy‑paste
• Archon OS: Advanced RAG with hybrid search and optional reranking (toggle in requirements)

Task Flow

• Traditional AI Coding Setup: Separate PM boards
• Archon OS: Task management integrated with knowledge and MCP tools

Feedback Loop

• Traditional AI Coding Setup: Manual
• Archon OS: Real-time updates via Socket.IO streams to UI and assistants

Scalability

• Traditional AI Coding Setup: Monolithic tools
• Archon OS: True microservices: UI, Server, MCP, Agents; independent scaling

Impact-oriented estimates developers can target with Archon’s architecture and workflow

Faster retrieval-to-answer for code context by consolidating sources and using semantic RAG.

Fewer integration misfires by standardising access via MCP tools.

Repository status: Public, active, community-driven; no binary releases yet listed.

Community activity: Growing GitHub Discussions for questions, ideas and announcements.

Positioning: “Command centre for AI coding assistants” with MCP server and advanced RAG; currently in beta.

Archon is a “true microservices” system with clear separation of concerns, HTTP boundaries and real-time streams to the UI:

Services and defaults:

• UI: React+Vite on port 3737
• Server (FastAPI): API, crawling, document processing, ML ops on port 8181
• MCP Server: lightweight HTTP wrapper with 10 MCP tools on port 8051
• Agents: PydanticAI agent hosting (streaming, reranking) on port 8052
• Backing store: Supabase (Postgres+pgvector)

Languages

• Python (server, MCP, agents), TypeScript (React UI)

Architecture

• Microservices: UI, Server, MCP, Agents; HTTP boundaries; Socket.IO streaming

Data Store

• Supabase (PostgreSQL + pgvector)

AI Providers

• OpenAI, Google Gemini and Ollama are configurable in Settings

RAG

• Hybrid search; contextual embeddings; optional reranking (enable in requirements)

MCP Tools

• 10 tools spanning RAG queries, task/project operations

UI Features

• Knowledge sources, crawl controls, uploads, projects, tasks, MCP dashboard

Real-time

• Socket.IO live progress and health checks

Ports/Config

• Configurable via .env for HOST and service ports

Prompt to Assistant (connected to Archon MCP): “Find and summarise the rate-limiting policy for API X from our docs.”

• Archon-Enabled Response Behaviour: Uses RAG tools to search the crawled knowledge base, reranks results, returns a concise summary with citations and attaches source links in context for follow-up

Prompt to Assistant (connected to Archon MCP): “Create tasks to implement OAuth with PKCE in our Next.js app.”

• Archon-Enabled Response Behaviour: Generates a feature with structured tasks tied to linked docs in the knowledge base; updates the project board and streams progress to the UI.

Prompt to Assistant (connected to Archon MCP): “Compare available embedding models configured in this workspace.”

• Archon-Enabled Response Behaviour: Reads current provider settings from Archon, summarises configurations and suggests alternatives if Ollama or Gemini is enabled.

• GitHub: Github Official Link - active repo with Quick Start, architecture diagrams.
• Discussions:  Q&A, feature requests, announcements; visible traction from users testing Docker, Supabase and MCP setups.
• Contribution pathway: Fork, branch, PR; project welcomes fixes/new features as it’s in active beta.

Prereqs:

• Docker Desktop
• Supabase account (free tier or local)
• API key for OpenAI (Gemini and Ollama supported)

Steps:

1. Clone

• git clone https://github.com/coleam00/archon.git
• cd archon

2. Configure

• cp .env.example .env
• Set SUPABASE_URL and SUPABASE_SERVICE_KEY (use the legacy long service key)
• Optional: enable reranking by uncommenting lines in python/requirements.server.txt (increases container size)

3. Database

• In Supabase SQL Editor, run migration/complete_setup.sql

4. Start services

• docker-compose up --build -d
• UI 3737, Server 8181, MCP 8051, Agents 8052 (configurable)

5. Configure providers

• Open http://localhost:3737 → Settings → choose provider and set API keys
• Test by crawling a site or uploading a doc

Reset if needed:

• In Supabase, run migration/RESET_DB.sql then re-run complete_setup.sql, restart containers, reconfigure keys and reupload sources.

Supported platforms:

• Docker-based deployment; UI runs locally in browser; backend services are containerised and portable where Docker is available.

Version: Beta (current)

• Date: Ongoing
• Feature: Microservices architecture; MCP tools; Knowledge base crawling and uploads; Hybrid RAG; Project/task integration; Socket.IO streaming

Q1. What is Archon in one sentence?: A command centre that centralises knowledge, context and tasks for AI coding assistants via an MCP server.

Q2. Does it work with Claude Code, Cursor, or Windsurf?: Yes, MCP-compatible assistants can connect and share the same context and tasks

Q3. How does Archon improve RAG quality?: It supports hybrid semantic search, contextual embeddings and optional reranking; it also extracts code examples from docs to enrich retrieval.

Q4. What data store do I need?: Supabase (Postgres+pgvector) is required today; migration SQL scripts are included.

Try Archon Today. Clone the repo, run the quick start and connect Claude Code or Cursor to experience AI coding with centralised context and task flow   

Join the community and shape the roadmap in Discussions Discussions On GitHub. 

Level up from junior to senior dev by engineering context, not just prompt code smarter, debug with drive and ship with speed.

Need help with AI transformation? Partner with OneClick to unlock your AI potential. Get in touch today!