io.github.helinakdogan/agent-magnet MCP Server
This MCP server provides a behavioral memory layer for AI tools. It learns preferences and context over time, addressing the issue where an AI forgets project details across new sessions. The server is described as sharing one memory across Claude, Cursor, and Codex.
π οΈ Key Features
- Learns automatically as you work (βno remember this requiredβ)
- Shares one memory across Claude, Cursor, and Codex
- Supports forgetting or marking items as done
π Use Cases
- Persisting project context across new sessions
- Maintaining consistent preferences while using multiple AI tools
- Marking completed information to avoid reusing it
β‘ Developer Benefits
- Cross-tool memory via a single shared memory layer
- Focus on context retention without manual memory commands
β οΈ Limitations
- Source data does not specify toolCount, supported platforms, configuration options, or memory storage details.