Adaptive Recall MCP Tool

[AIAppsAPI]

Pre-submission Checklist

• This post follows the MCP community guidelines

What would you like to share?

Adaptive Recall is a hosted memory server that connects via MCP or REST API. It stores memories across six typed categories, retrieves them using four parallel strategies (vector, temporal, keyword, and knowledge graph), and ranks results using ACT-R activation modeling from cognitive science research. The system trains ML models on your usage patterns to optimize retrieval. Every ML driven parameter change must pass a statistical significance test against real query history before being accepted. Memories have a lifecycle with confidence evolution, automatic knowledge graph construction, retrieval quality self verification, and curiosity driven knowledge gap detection.

I found you trying to submit to your MCP directory, hoping for inclusion please. There is a free tier which is plenty to test with.
Thank you,
Paul

Relevant Links

https://github.com/AIAppsAPI/adaptive-recall
https://www.adaptiverecall.com/

[armorer-labs]

Memory and recall tools feel especially important from a safety perspective because stored context can quietly become a long-lived trust channel.\n\nA lot of systems think carefully about user prompts and tool calls, but less about what happens when untrusted or partially trusted text gets written into memory and later comes back with a kind of accidental authority.\n\nSo one question I would ask of any recall layer is: can it preserve provenance strongly enough that retrieved memory is still treated according to where it came from? User note, tool result, scraped page, approval artifact, policy text, etc. should probably not all come back with the same weight.\n\nThat provenance question seems just as important as retrieval quality once agents start making decisions based on stored context.