MCP servers that ship as code — a repository or a package you install and run locally (usually over stdio). They have no hosted endpoint, so we cannot probe or grade them: instead each one shows its repository, install command, license, stars and when it was last updated.
Collected from GitHub repositories and npm, PyPI, Cargo, Go, Maven, NuGet and RubyGems packages tagged as MCP servers, the official MCP Registry, Docker's MCP catalog, ModelScope, Baidu MCP World, Cline and other public directories; one entry per repository. A server that also offers a hosted endpoint is listed under Remote instead. These entries are not counted in our reachability, verification or readiness statistics.
-
cozo-memorygithub.com/tobs-code/cozo-memoryupdated this year
Persistent, local-first memory for AI agents. No cloud, no Docker, no external services – just CozoDB embedded in Node.js.
npm Apache-2.0 ★ 1 -
mementogithub.com/wmyung/mementoupdated this year
Agent memory: SQLite fact store + Git wiki + keyword bridge. Multi-agent memory system for AI agents. No embeddings, no vector DB. Zero-dependency Python CLI with MCP server.
MIT -
Search your C# codebase by meaning, not keywords. 100% local - Ollama embeddings + SQLite vectors + MCP server. No cloud, no telemetry.
MIT ★ 7 -
Semantic code search for AI agents and CLI users. CLI + MCP server using local ONNX embeddings — no API keys, no external services.
nuget ★ 1 -
cognee-mcp-fastembedgithub.com/Lauredmarin/cognee-mcp-fastembedupdated this year
cognee-mcp with local fastembed embeddings (no external embedding API)
-
mcp-mozartgithub.com/gunesbizim/mcp-mozartupdated this year
Recommender MCP server. Auto-discovers other MCPs registered in a Claude Code project and recommends which tools to consult per query, using local embeddings. No fan-out, no per-query LLM call.
npm MIT -
Long-term memory for AI agents: hybrid-searched memories over human-editable markdown notes. 100% local — offline ONNX embeddings, SQLite FTS5, MCP server, REST API, CLI, dashboard. No cloud, no API keys.
npm ISC -
MCP server that hands coding agents the one Markdown section that answers the question, not the whole file. Local hybrid search: BM25 keywords + ONNX embeddings in SQLite. No API keys, no network.
MIT ★ 1 -
semantic-search-mcpgithub.com/zulaxy/semantic-search-mcpupdated this year
Semantic code search MCP server for AI coding agents (opencode, Claude, Cursor). Local embeddings, no API keys.
npm MIT -
Embedded single-file store for facts that change: exact vector search, full revision history, no server. numpy-only.
pypi Apache-2.0 ★ 1 -
@adamrdrew/agent-memory-mcpgithub.com/adamrdrew/agent-memory-mcpupdated this year
MCP server for persistent agent memory — hybrid BM25 + vector search backed by LanceDB with local ONNX embeddings. No API keys, fully local.
npm GPL-3.0-or-later -
agidbgithub.com/rohansx/agidbupdated this year
A cognitive substrate for AI agents — hyperdimensional memory with first-class goals, beliefs, sensory, self-model, and a non-destructive unlearn primitive. Rust top to bottom, embedded-first, no LLM in the read path. One binary, no query language.
Apache-2.0 ★ 6 -
A context-first, deterministic, single-file database for AI agents — SQLite for AI memory. Records, typed graph relations, BYO embeddings, and time under one ACID transaction. No-LLM by contract: hybrid kNN+BM25, MATCH/CLOSURE traversal, AS OF time travel, NQL grammar, MCP server included.
Apache-2.0 ★ 2 -
descrygithub.com/jKiler/descryupdated this year
Local hybrid (vector + BM25) code search for AI coding agents — AST chunking, MiniLM embeddings on ONNX, Go call graph, MCP server. No external services.
go Apache-2.0 ★ 1 -
mcp-recallgithub.com/YanpengQi7/mcp-recallupdated this year
Local-first structured memory for Claude Code over MCP — semantic recall of incidents, code reading, decisions, and lessons. SQLite + sqlite-vec, local embeddings, no API keys.
★ 1 -
Long-term memory your AI coding agents actually share. One brain, many agents — standalone embedded MCP server: SQLite + FTS5 + vector hybrid search, local ONNX embeddings, self-expiring diary. No server, container, or daemon.
pypi MIT ★ 3 -
Personal memory for Claude Code and Codex. Save links, notes, and files from inside the agent; find them later by describing what you remember. Local SQLite + embeddings, no API key.
npm MIT -
needlegithub.com/alexandremourin/needleupdated this year
Semantic code search MCP server: tree-sitter chunking + local Model2Vec embeddings + BM25/dense hybrid (RRF). Hands an AI only the relevant code (file:line), ~10-100x fewer tokens. Multi-language, no API key.
MIT -
lean-memorygithub.com/wuesteon/lean-memoryno update data
Embedded, local-first agent memory: one SQLite file per namespace, no server, no cloud.
pypi -
Truly Zero-Dependency Advanced MCP server with local file-based memory features for VS Code/Cursor - NO EXTERNAL API REQUIRED - Built-in AI embeddings using native algorithms - Zero-config local semantic search (Embedding v1.2.0, FileSystem v2.0.0, Memory
npm MIT -
Long-term memory for AI agents: one SQLite file, local embeddings, hybrid recall (sqlite-vec + FTS5), rules-based consolidation into facts and procedures. MCP server + CLI, no cloud.
MIT -
Local-first operating memory for AI coding agents - Claude Code, Codex, and any MCP client. Tracks decisions and their evolution. SQLite + on-device embeddings, no cloud.
go MIT ★ 15 -
Long-term memory for AI agents as plain JSON objects in S3 (or R2, SeaweedFS, MinIO, a directory): no database, BM25 + local embeddings, conditional-write concurrency, MCP server, LangGraph store.
MIT -
rmbrgithub.com/srock44/rmbrno update data
Embedded, local-first memory and retrieval for AI agents. One SQLite file, no server, no API key.
pypi -
Local MCP server for semantic code search. AST-aware chunking with tree-sitter, offline ONNX embeddings, hybrid vector + BM25 (RRF) ranking, one SQLite file. No API keys or cloud. Semantic code search for Claude Code, Cursor, opencode and AI coding agents.
npm MIT -
scout-code-navigatorupdated this year
MCP server for fast, incremental code navigation - no indexing, no embeddings, no waiting
pypi MIT -
hexusgithub.com/codenamekt/hexusupdated this year
Hexus: Postgres + hexus memory provider plugin for hermes-agent AND a standalone MCP server. Local BERT (MiniLM-L6-v2) embeddings replace the upstream HTTP embedder; multi-agent storage with per-minion themes, async writer, no LLM in the memory hot path.
pypi BSD-3-Clause -
Give your coding agent the right code, not more code. Local context engine for Claude Code, Cursor & Codex: code graph + hybrid search + embeddings + git history, packed into a token budget. Held-out SWE-bench: right file in 4k tokens 92% vs 29% for search + whole files. MIT, no API keys.
pypi MIT -
zakira.exchangegithub.com/moaidhathot/zakira.exchangeupdated this year
An MCP server and CLI tool for AI agent memory storage and semantic search. Enables agents to save, share, and search memories using hybrid search (BM25 + vector embeddings via all-MiniLM-L6-v2), merged with Reciprocal Rank Fusion. Built with .NET 10 and SQLite -- standalone with no external service dependencies.
nuget Unlicense -
codebase-cortexgithub.com/sarupurisailalith/codebase-cortexupdated this year
AI-powered documentation autopilot — commit code, docs update themselves. MCP server mode gives coding agents (Claude Code, Cursor, Windsurf) 11 documentation tools with no LLM key needed. Also runs as a standalone local-first pipeline with FAISS embeddings and section-level doc updates. Works with any LLM, optional Notion sync.
pypi MIT -
MCP server for Revit API documentation and The Building Coder blog. No API keys, no embeddings, Revit not required.
MIT -
memfabricgithub.com/MrBoor/memfabricupdated this year
Self-organizing agent memory. No embeddings, no vector DB.
MIT ★ 1 -
mcp-local-memorygithub.com/NickSmet/mcp-local-memoryupdated this year
Lightweight local memory for AI agents. SQLite + embeddings, zero setup, no services to run. Optional OpenAI support.
npm MIT -
portable-memory-mcpgithub.com/slim00walid-prog/portable-memory-mcpupdated this year
One memory that follows you across every AI app — a user-owned MCP memory store shared by Claude, ChatGPT & Cursor. No cloud, no embeddings, no lock-in
MIT -
amnitexgithub.com/amnibro/amnitexupdated this year
atex (amni-tex): a lossless byte-page memory layer for MCP-capable AI coding assistants. Local, no embeddings, no cloud.
pypi MIT -
Vectorless agent memory. HTML storage + grep retrieval. LongMemEval-S R@5 = 98.9%. No embedding, no vector DB.
MIT -
Local-first RAG server for Verse and UEFN docs. Ollama embeddings, pgvector storage, MCP interface, fully air-gapped with no cloud calls
MIT ★ 1 -
📚 Any ebook → clean token-counted Markdown chapters. Let Claude read your bookshelf over MCP — and catch it misquoting your books. Deterministic RAG: no embeddings, no network, no API keys.
pypi MIT ★ 2 -
llm-wikimindgithub.com/HAL-9909/llm-wikimindupdated this year
The production-ready implementation of Karpathy's LLM Wiki pattern. Markdown + BM25 + MCP. No embeddings, no vector DB.
MIT ★ 2 -
Browserless carousel factory — 46 slide types, embedded Blitz renderer (no Chromium), MCP server, WCAG-AA validation. 185 tests.
MIT -
KnowledgeGithubMCPgithub.com/soynx/KnowledgeGithubMCPupdated this year
MCP server that turns any GitHub repo into a AI knowledge base — no embeddings, no vector DB, one Docker container.
MIT -
Local caching/memory layer for Claude Code — avoids re-spending tokens on identical or near-duplicate queries. 100% local: SQLite + local embeddings, no telemetry, no external calls.
npm MIT -
A memory for AI assistants that is just a SQLite file: full-text search over MCP. No vector store, no embeddings, no external service.
pypi MIT ★ 1 -
Drive ArmorPaint 1.0 from an MCP client: 58 tools over a bridge plugin written in the app's own embedded C dialect. No fork, no patched binary required.
NOASSERTION ★ 4 -
mneogithub.com/HugoLopes45/mneoupdated this year
Persistent memory for AI agents using git refs. MCP server + Claude Code skill. No vector database, no embeddings, no daemon — branches are scopes, push/fetch is sync.
MIT ★ 2 -
OpenContextScitoolgithub.com/aminHZDEV/OpenContextScitoolupdated this year
LLM context handler over an Open Knowledge Format bundle BM25 search via MCP, usage telemetry, and a health dashboard. No embeddings, no API key.
pypi -
Agent-native code intelligence: turn any repo into a queryable code graph over MCP. tree-sitter + SQLite, no embeddings, no API keys.
MIT -
only-my-mem0rygithub.com/ost527/only-my-mem0ryupdated this year
A fully local, zero-config Mem0 memory server for MCP clients on macOS — no LLM, no API key, no cloud. The client is the brain; on-device embeddings + Chroma, auto-managed launchd lifecycle.
MIT ★ 6 -
A fully local MCP server providing AI agents with persistent semantic long-term memory using SQLite and transformer-based embeddings. No cloud dependencies or API keys required.
npm MIT ★ 2 -
witsloggithub.com/all-wits/witslogupdated this year
AI-native error intelligence framework — structured error events in a local SQLite DB per project, full-text searchable, deterministically auto-classified, exposed to any MCP-compatible AI assistant. No embeddings, no cloud, no infra.
Apache-2.0 ★ 1 -
Persistent, cross-runtime, typed memory for AI agents — MCP stdio server over embedded SQLite. Local, offline, no cloud.
MIT -
factorforgegithub.com/aman9824/factorforgeupdated this year
Vectorless retrieval (no embeddings) over an evidence-linked knowledge graph, served as a standalone MCP server. Multi-agent Claude research pipeline.
MIT -
json-mcp-liteno update data
Turn any JSON file (product catalogue, FAQ, price list) into a read-only MCP server. Exposes list, search and get tools over stdio. Keyword search, no embeddings, no API key, two dependencies. Works with Claude Desktop, Claude Code and Cursor. MIT.
smithery -
A code graph an LLM can actually use. Tree-sitter across 6 languages, typed ontology with inferred architectural roles, impact analysis, 9 interactive views, MCP server. No embeddings, no vector DB.
MIT ★ 2 -
Durable, file-based long-term memory for AI agents. Five-package plugin family: SDK, CLI, MCP server, Hermes adapter, and a LangGraph BaseStore. No vector database, no embeddings.
pypi MIT ★ 177 -
prune-mcpgithub.com/dev7shah/pruneupdated this year
Context-selection engine for AI coding assistants — finds the most relevant files in a codebase for a change request. No LLM, no embeddings, pure graph + text scoring.
npm MIT -
docs-indexgithub.com/mohsingdp-ai/print-press-docsupdated this year
A Cursor-style @Docs for the terminal: index any docs site into a local SQLite + FTS5 store, then search, read, and serve it to Claude over MCP — no re-fetching, no embeddings provider.
npm MIT -
Component lookup over the encode-ui shadcn registry, as an MCP server. Zero setup; local embeddings, no API keys.
npm MIT -
code-indexgithub.com/hdimer/code-indexupdated this year
Semantic code search as an MCP server. Find code by meaning, not keywords. Local embeddings + Milvus Lite, no API keys.
MIT -
traceon-cligithub.com/dhanushmurugesang/traceonupdated this year
Runtime verification for AI coding agents. MCP server that lets Claude Code verify code changes work end-to-end via Playwright + OpenTelemetry. Embedded OTel collector by default — no Docker, no SigNoz required.
npm MIT