Midas
Local-first, source-traceable agent memory with no LLM at ingest, running fully offline.
Midas is a local-first memory system for AI agents that stores and retrieves information without requiring an LLM at ingest time. It uses embedding-based recall, importance scoring, and belief revision to maintain a SQLite-backed memory store. The MCP server exposes tools for storing, recalling, capturing, and forgetting memories, with support for multiple embedding backends and offline operation.
Source
Repository: https://github.com/vornicx/midas
Maintain Midas?
Let people know it's listed here — add the badge (live metrics, light/dark aware) or a plain link to your README or docs.
[Midas on getagentictools](https://getagentictools.com/mcp/vornicx-midas?ref=badge) Related mcp
aaPanel Bridges AI with aaPanel web hosting control panel for managing websites, databases, Docker containers, and system resources throug
CockroachDB Integrates with CockroachDB to provide cluster monitoring, database operations, table management, and SQL query execution with tra
Gel Database Integrates with Gel databases to enable automated schema discovery, query validation, and execution of EdgeQL and TypeScript queri
MySQL Integrates with MySQL databases, enabling SELECT queries, table listings, and schema retrieval for data analysis and dynamic datab
OpenFDA Semantic Intent-driven access to the full openFDA API ecosystem covering drugs, devices, food, and adverse event reports.
Simple PostgreSQL Provides direct SQL query execution against PostgreSQL databases with configurable read-only and write operation modes, connection