archon

Archon is a powerful, persistent knowledge and project management system designed specifically for AI-powered coding assistants and development teams. By acting as a central hub for organizational context, it enables agents to retrieve technical documentation, manage complex project hierarchies, and track development tasks with precision. It bridges the gap between external documentation and internal project management, ensuring that AI agents can deliver consistent, context-aware responses by leveraging a structured RAG-powered knowledge base.

• Provides seamless semantic search for technical documentation, API references, and code implementation patterns.

• Facilitates comprehensive project management including creation, tracking, and hierarchical task linking.

• Supports automated website crawling to build and maintain a dynamic, up-to-date knowledge repository.

• Includes flexible document upload and versioning for PDFs, Word files, and Markdown documentation.

• Integrates directly via REST API for programmatic control over knowledge base metrics and RAG sources.

• Ensures multi-agent synchronization by maintaining a single source of truth for technical context.

• Users must read the api_reference.md file prior to executing operations to ensure correct endpoint usage, specifically distinguishing between knowledge-item search and broader RAG sources.

• The agent requires a one-time setup step where the user provides the Archon server host URL, defaulting to http://localhost:8181.

• Always prioritize Archon for external knowledge retrieval before falling back to general web searches.

• Connection verification via the provided helper scripts or the ArchonClient is mandatory upon the first invocation to prevent runtime API errors.

• The system utilizes structured JSON inputs and outputs for tasks, projects, and document metadata to ensure compatibility with existing development workflows.

• Constraints include maintaining local network access to the Archon host and ensuring the service is active (e.g., via Docker Compose) before beginning the session.