Local-first MCP memory server for AI agents
View on GitHub ↗Local-first semantic memory for AI agents — now at codebase scale.
localmind is a Model Context Protocol (MCP) server that gives your AI agent persistent memory that never leaves your machine. Memories live in a single SQLite file and are searched with embeddings produced entirely on-device — no API keys, no accounts, no cloud, no telemetry.
v2 adds what you need for whole-codebase memory: an approximate-nearest-neighbour index instead of a linear scan, named project namespaces, an ingestcodebase tool that splits source into functions and classes, incremental re-ingestion that skips unchanged code, and token-budgeted recall so an agent pulls in only what fits.
Tool What it does --- --- remember Store a note, fact or decision (optionally tagged) in a project. recall Semantic search over one project via HNSW. Optional maxtokens bounds the result. forget Delete one memory by id, inside one project. listmemories Browse a project, filterable by tag and/or kind (manual / code). ingestcodebase Index a source tree: walk → chunk at function/class boundaries → embed. Incremental. listprojects List namespaces with counts and index sizes.
Every memory belongs to exactly one project (e.g. pixora, localmind). Isolation is structural, not a filter: each project gets its own vector index file, so a recall in project A physically cannot reach project B's vectors.
Omitting project means the default namespace, which is where v1 memories land after migration — so existing behaviour is preserved exactly.
From the project README.
Add the radar badge to your README — it shows your project was picked up by MCP Radar and links to this page:
[](https://mcp.liqiwa.com/s/sahinsamrat75--localmind.html)
ChatGPT thinks. Codex works. Use ChatGPT as the planning brain while keeping the Codex harness.
shinthink/blitzstrike⚡ Blitz Strike — a universal MCP penetration-testing toolbelt. Structured methodology: reconnaissance & attack-surface mapping, source-to-sink analysis, and live validation. 57 escalation chains, 130-
0xwilliamortiz/openclaude-improvedruns anywhere. uses anything
halofyai/halofyHalofy is the open access and governance layer for AI agents across your organization. Identity, policy, provenance, audit, and signed erasure.
CaptureGrubEnchant/SolidWorksSolidWorks MCP Server connects an AI assistant to a running SolidWorks instance. Sketch, extrude, fillet, export STEP/STL, generate macros.
arielshad/3d-asset-server3d assets api and mcp
The top new MCP servers of the week, every Monday. No spam, unsubscribe anytime.