Home › Knowledge & Memory › Midas
Midas
Local-first, source-traceable agent memory — no LLM at ingest, fully offline
Topics: Knowledge & Memory
Connect
Review any command before running it. Package names and URLs come from the server's own registry entry.
Package (pypi 0.0.1)
uvx midas-memory-mcp
Or add to your MCP client config:
{
"mcpServers": {
"midas": {
"command": "uvx",
"args": [
"midas-memory-mcp"
]
}
}
}
MIDAS_MCP_DB— Path to a SQLite file to persist memory across restarts (default: in-memory)MIDAS_MCP_EMBEDDER— Embedding backend: 'local' (bge ONNX, offline) or 'hashing' (default: local)MIDAS_MCP_MAX_RECORDS— Cap the store; above it the lowest-value memories are auto-forgotten (no LLM)MIDAS_MCP_MIN_IMPORTANCE— Relevance floor 1-5 for auto-capture; turns scoring below it are skipped
Related servers
modelcontextprotocol/servers
Direct local file system access.
Context7
Up-to-date docs for AI agents
hindsight
Hindsight: long-term memory for AI agents modeled on human memory.
cognee
Memory manager for AI apps and agents using graph and vector stores, with ingestion from many data sources.
Skill_Seekers
Turn docs, GitHub repos, PDFs, videos, Confluence, Notion and Slack/Discord into AI skills and RAG knowledge, exportable to vector…
basic-memory
Local-first knowledge management with bi-directional LLM sync via Markdown files.
Listed in punkpeye/awesome-mcp-servers (MIT)
Data from the Official MCP Registry