The OFFICIAL Archon Guide - 10x Your AI Coding Workflow

Key Concepts

Archon: A command center for AI coding designed to facilitate deep human-AI collaboration by managing knowledge and tasks for AI coding assistants.
MCP Server: Archon functions as an MCP server, allowing easy connection and communication with virtually any AI coding assistant.
RAG (Retrieval Augmented Generation): A powerful strategy employed by Archon for both keyword and semantic searching of curated documentation, enabling AI assistants to access up-to-date and relevant information.
Superbase: The underlying database technology powering Archon, offering a free cloud tier and local hosting options.
LLM (Large Language Model) Providers: Archon supports various LLM providers like OpenAI, Gemini, and Olama, allowing users to choose their preferred models for chat and embeddings.
Training Cutoff: The inherent limitation of LLMs regarding their knowledge of recent libraries or frameworks, which Archon addresses through RAG.
llm.ext & sitemap.xml: Optimal file formats for crawling and ingesting documentation into Archon's knowledge base.
Global Rules: Configuration settings for AI coding assistants that instruct them on how to interact with Archon for knowledge and task management.
Kanban Board: A visual task management system within Archon that allows real-time collaboration between humans and AI on project tasks.

Introduction to Archon

Archon is presented as a comprehensive tool designed to be the central command center for AI coding, bridging a significant gap in current AI coding practices by enabling deep collaboration between AI and human developers. It offers a user-friendly interface for managing projects, tasks, and curated knowledge (codebases, libraries) for human users, while simultaneously acting as an MCP server for AI coding assistants to access and manage the same resources. This dual functionality allows for seamless integration and real-time updates, putting the user in control while empowering AI assistants.

The video outlines three main objectives:

Explain the necessity and benefits of using Archon.
Provide a quick start guide for setting up Archon and connecting it to an AI coding assistant.
Offer a practical guide on integrating Archon into AI coding workflows to enhance code quality and management.

Why Archon is Needed: Addressing AI Coding Assistant Limitations

The core problem Archon aims to solve is the lack of collaborative control with existing AI coding assistants (e.g., Kirao, Cloud Code, Codeex). While these tools are powerful, they often operate too much "under the hood":

Uncontrolled Web Search: They search the entire internet for external documentation (e.g., LangChain, Pydantic AI), preventing users from specifying or curating the knowledge sources.
Internal Task Management: They create and manage their own internal task lists, which are not easily accessible or interactive for human collaboration.

This "kicks us out of the driver's seat," as the presenter states, hindering the ability to define tasks or choose specific knowledge sources for the AI. Archon addresses this by allowing users to:

Curate Knowledge: Point the AI to specific, pre-curated documentation.
Collaborate on Tasks: Interact with a shared task list (Kanban board) in real-time.

Archon's Core Functionalities and Demo

Archon's primary functionalities revolve around knowledge and task management:

Knowledge Management

User-Curated Documentation: Users can direct Archon to specific documentation sources.
RAG Implementation: Archon performs powerful RAG (Retrieval Augmented Generation) using both keyword and semantic searching on the curated knowledge base.
Real-time Access: Once documentation is crawled, AI assistants immediately gain access to search through it via the MCP server.

Task Management

Shared Kanban Board: Archon provides a Kanban-style board for managing projects and tasks.
Real-time Updates: AI assistants can move tasks between stages (e.g., "To Do," "Doing," "Review") in real-time, with updates visible in the UI without refreshing.
Human-AI Collaboration: Users can create new tasks or edit existing ones, and the AI assistant will incorporate these changes into its workflow.

Demo Example: The video demonstrates an AI assistant moving tasks on a Kanban board from "To Do" to "Doing" and then "Review" in real-time, simulating an AI development workflow. The presenter also highlights future plans for Archon, including background tasks for planning and validating pull requests.

Setting Up Archon: A Quick Start Guide

Setting up Archon is designed to be straightforward, with most steps guided through the user interface.

Prerequisites

Docker or Docker Desktop: Archon runs as several containers.
Superbase: The database powering Archon.
- A free tier for cloud Superbase is available and used in the demonstration.
- Local Superbase hosting is also an option.
- Requires Superbase URL and Service Role Key (found in Project Settings > Data API and API Keys > Service Role Secret, respectively). For local Superbase, the URL is typically host.docker.internal:port and the service role key is self-set.
LLM Provider API Key: Supports OpenAI, Gemini, and Olama (for 100% local, private knowledge bases).

Step-by-Step Setup Process

Clone the Archon Repository:
- git clone https://github.com/archon/archon.git
- Use the stable branch for official releases (or main for experimental features).
- Change directory into archon.
Set Up Environment Variables:
- Copy .env.example to .env.
- Set SUPERBASE_URL and SUPERBASE_SERVICE_ROLE_KEY in the .env file. Other keys (like OpenAI API key) are set in the UI.
Set Up the Database:
- Navigate to migrations/complete_setup.sql in the cloned repository.
- Copy the SQL content.
- In the Superbase dashboard, go to the SQL Editor, create a new snippet, paste the SQL, and run it. This creates all necessary Archon tables.
Start Archon Containers:
- In the terminal within the archon folder, run: docker compose up --build -d
- This builds and starts containers for the user interface, MCP server, and API endpoints. It may take 10-15 minutes.
Configure Archon User Interface:
- Access Archon UI at localhost:3737 (default port).
- Go to the Settings page.
- Enter credentials for desired LLM providers (e.g., OpenAI API key).
- Select preferred chat and embedding models (e.g., OpenAI for chat, Olama for embeddings). Save settings for both.
Connect to AI Coding Assistant (MCP Tab):
- Go to the MCP tab in Archon UI for instructions specific to various AI coding assistants (e.g., Cloud Code).
- Copy the provided command (e.g., claude mcp connect) and paste it into your AI coding assistant's terminal.
- Verify connection (e.g., claude mcp list).
Set Up Global Rules:
- In Archon settings, copy the recommended global rules for your coding assistant (e.g., for Cloud Code, paste into cloud.md; for Codeex, into agents.mmd).
- These rules teach the AI assistant how to use Archon for knowledge and task management.

Upgrading Archon

git pull to get the latest changes.
Run docker compose up --build -d again.
Check the settings page for any database migrations; copy and run the provided SQL in the Superbase SQL editor.

Superbase Sponsorship & Scalability

The video includes a sponsored segment by Superbase, highlighting its role as the database powering Archon. Key takeaways from the Superbase Select conference mentioned are:

Remote MCP Server: New feature allowing agents to manage Superbase projects without local installation.
Scalability Enhancements: Incorporating S3 buckets, acquiring "Oreo," and a new open-source initiative called Multi-gress for database orchestration, aimed at scaling Postgres to millions of users.
Series E Funding: Superbase recently secured $100 million, valuing the company at $5 billion.
Technical Deep Dive: A talk by Sugamarine (linked in description) provides more technical details on Multi-gress and Superbase's scalability architecture.

Practical Usage in AI Coding Workflows

The presenter emphasizes that users have full flexibility to define their own AI coding workflows and how Archon integrates.

1. Setting Up Knowledge for a New Project/Feature

Identify Resources: Determine which files or documentation the AI assistant needs to reference (e.g., Pydantic AI, Superbase SDK).
Crawl Documentation:
- Go to Archon UI > "Add New Knowledge."
- Paste the URL of the documentation.
- Optimal Formats: Prioritize llm.ext (e.g., superbase.com/docs/llm.ext) or sitemap.xml (e.g., memzero.com/sitemap.xml) for better ingestion as markdown.
- Archon can crawl multiple sources in parallel.
Immediate Access: Once crawling is complete, the AI assistant can immediately search through the private knowledge base via MCP.

2. Integrating Archon into Planning and Implementation Workflows

The presenter advocates for splitting coding into planning and implementation phases, using reusable markdown documents as workflows.

Planning Phase Example

Reusable Workflow: A markdown document outlining steps like "Read and analyze requirements," "Research phase," "Codebase analysis," and "Plan and design."
Archon Integration:
- The "Research phase" explicitly tells the AI to list sources crawled in Archon and perform RAG to look through chunks and code examples.
- Example Request: Building a Pydantic AI agent that uses Memzero, requiring Archon to leverage both Pydantic AI and Memzero documentation.
Output: A detailed implementation plan with research findings from Archon and a list of tasks.

Implementation Phase Example

Reusable Workflow: A markdown document focusing on leveraging Archon's project and task management.
Archon Integration:
- Instructs the AI to check global rules for existing Archon projects.
- If no project is found, it creates a new one.
- Manages tasks by creating new Archon tasks for each implementation step and moving them through the Kanban board.
Real-time Task Management: During execution, the AI creates a new project in Archon, converts plan tasks into Archon tasks, and moves them (e.g., "Set up initial project structure" to "Doing"). Users can view task descriptions, edit instructions live, or add new tasks.
Improved Code Quality: The AI, by leveraging Archon's RAG on specific documentation (e.g., Memzero, Pydantic AI), produces accurate code that it wouldn't know due to its training cutoff.

The presenter encourages users to customize these workflows and take inspiration from the provided examples (linked in the description as slash commands/prompts). Archon is compatible with any AI coding assistant, not just Cloud Code.

Conclusion

Archon is positioned as an essential tool for modern AI coding, fundamentally changing how developers collaborate with AI assistants. By providing a centralized platform for knowledge curation and collaborative task management, Archon addresses the limitations of current AI tools, putting human developers back in the driver's seat. Its easy setup, flexible integration into custom workflows, and powerful RAG capabilities significantly improve the quality and efficiency of AI-generated code, especially when dealing with new or specialized libraries. The video concludes by promising more content on integrating Archon into end-to-end projects.