ATLAS (Adaptive Task & Logic Automation System) is a project, knowledge, and task management system for LLM Agents.
Built on a 3-node architecture:
MCP.Pizza Chef: cyanheads
ATLAS gave an assistant somewhere to keep its work: projects at the top, tasks underneath, and notes or research attached to either, all searchable together. Later versions added a deep research mode. Two things should give you pause. It stopped moving, with the last change in June 2025, more than a year ago. And it was never easy to run, because it needs a Neo4j database of your own, either through Docker or a paid cloud account. Find a maintained alternative before you commit.
Unmaintained · no updates since June 2025; no direct replacement named
Data
Notes
Use This MCP server To
Keep a running list of projects my assistant can update Break a project into tasks with priorities and deadlines Attach research notes to the project they belong to Search across projects, tasks and notes at once Track which tasks are blocked by other tasks
README
atlas-mcp-server FAQ
Is this project still maintained?
No. The last change was in June 2025, over a year ago. It has not been archived, but no fixes or releases have appeared since. The same author publishes many newer connectors that are updated constantly, though none is a direct replacement for this one.
What do I need to run it?
A Neo4j database of your own, either self-hosted with Docker or a paid Neo4j cloud account, plus its sign-in details. Version 1.5.4 was the last one that used a simple local file instead.
Which apps does it work in?
It was built for Claude Desktop and other assistants that support MCP.
Can I use this to keep track of a long project?
In principle yes, since projects, tasks and notes are all supported. But with no updates in over a year, a maintained task tool is the safer bet.
How hard is setup?
Hard. There is a database to install and keep running, plus configuration by hand. This is not a paste-and-go connector.
Is my data stored anywhere I do not control?
Only if you choose the paid cloud database. Run Neo4j yourself and everything stays on your own machine.