Detailed Introduction
ATLAS is an MCP (Model Context Protocol) server built on Neo4j that provides structured project, task and knowledge management capabilities for LLM agents. It implements a three-tier entity model (Projects, Tasks, Knowledge) and offers unified search, dependency management and citation-ready knowledge items, enabling agents to operate within auditable and orchestrated workflows. ATLAS can be self-hosted via Docker Compose or connected to cloud Neo4j services such as Neo4j Aura.
Main Features
- Project and task lifecycle management with bulk operations, dependencies and priority handling.
- Knowledge repository integration that links knowledge items to projects and tasks, enabling RAG-style workflows.
- Graph-powered architecture using Neo4j for native relationship modeling and efficient cross-entity queries.
- Multiple transport modes (stdio and streamable HTTP) for easy integration with MCP clients like IDE extensions.
- Built-in Deep Research tooling to generate hierarchical research plans and export examples.
Use Cases
- Orchestrating LLM agents to automate task generation, assignment and progress tracking.
- Structuring research workflows (Deep Research) with hierarchical plans and knowledge citations.
- Team collaboration and project tracking with improved contextual traceability.
- Building RAG or retrieval layers backed by a graph database to exploit richer context links.
Technical Features
- Protocol compatibility: implements MCP for interoperability within the MCP ecosystem.
- Stack and implementation: written in TypeScript/Node.js with modular layout (src/mcp, services, utils).
- Data and backup: includes database backup and restore scripts for operational reliability.
- Extensibility: exposes atlas_* tools for programmatic automation and integration.
- Open source: Apache-2.0 licensed, with examples and documentation available in the repository.