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.
-
spath-enginegithub.com/shamik-13/spath-engineupdated this year
Semantic path engine for AI coding agents — resolves spath addresses to source code via MCP
npm Apache-2.0 -
Local-first AI/ML research curation: Quick Search, Learning Path, Research Lineage, Trending Feed, My Feed, Venues. Data from Semantic Scholar, HuggingFace, GitHub. MCP server included.
MIT ★ 2 -
Code intelligence MCP server for AI coding agents — local semantic code search, symbol/call-graph analysis (communities, dead code, path tracing), and persistent agent memory, served over stdio and Streamable HTTP.
MIT -
smart-connections-mcpgithub.com/gogogadgetbytes/smart-connections-mcpupdated this year
Secure, read-only MCP server for Obsidian Smart Connections. Path-validated semantic search for Claude Code.
npm MIT ★ 3 -
Local-first MCP server for searching private Obsidian vaults with hybrid full-text, fuzzy, semantic, and wikilink graph retrieval. Runs locally via Node/npm; only the vault path is required.
smithery -
Advanced ADB MCP server for Android automation — optimised for small LLMs (gpt-oss-20b, qwen3vl). 15 semantic tools, OCR + Set-of-Mark dual-path screen understanding.
MIT ★ 2 -
Yggdrasilgithub.com/IT-Square-Plus/Yggdrasilupdated this year
MCP memory server for Claude Code with semantic search. Chroma Cloud storage, ONNX embeddings, path-based routing. One container serves multiple projects with isolated memory spaces. FastMCP + FastAPI.
MIT ★ 1 -
reponovagithub.com/CristianoCiuti/reponovaupdated this year
Turn your codebase into a knowledge graph. 11 MCP tools for AI code assistants — search, impact analysis, shortest path, semantic similarity, and more.
npm MIT ★ 8 -
DevAssistant-MCPgithub.com/p3aRA55/DevAssistant-MCPupdated this year
A secure MCP server bridging AI coding assistants to local dev workflows — file I/O with path sandboxing, GitHub integration, pytest automation, and semantic codebase search via ChromaDB.
-
Microsoft Fabric enterprise BI reference build: hot and cold paths, medallion Lakehouse, semantic model, governed data agent (MCP + A2A), Purview-style governance, Terraform + Bicep. Runs offline on synthetic data.
MIT -
TalkDBgithub.com/Nitin-Gupta1109/TalkDBupdated this year
MCP-native autonomous data analyst. Natural language → validated SQL, with multi-turn conversation, statistical insight agent, proactive watchdog, dual-path verification, and a community registry of semantic models.
MIT ★ 1 -
MCP server for deep semantic analysis of Scala via SemanticDB — exact find-usages, class hierarchies, implicit resolution & call paths for AI coding agents like Claude Code. Beyond grep and standard LSP.
maven MIT ★ 23 -
claude-research-packgithub.com/The-Fish-Salmon/claude-research-packupdated this year
Portable Claude research pack for Windows: three install paths (Code+WSL, Code-native, Desktop). Deep-research, paper-capture into Obsidian, lit-status, MCP servers (arxiv/Semantic Scholar/scihub/university-paper-access/obsidian), and v3 cross-device research-continuity skills.
NOASSERTION ★ 3 -
Local graph memory for AI agents. Every answer comes with the path it followed. Stores memories in SQLite, links the people, projects and tools they mention, and combines semantic search with graph traversal. Works as a CLI, MCP server and local web viewer.
MIT ★ 2 -
A branch-aware, dual-revision epistemic ledger and control plane for AI coding agents. From requirements and isolated work to validated changes and durable, queryable project truth.
MIT ★ 40 -
markdown-fastrag-mcpgithub.com/lidge-jun/markdown-fastrag-mcpupdated this year
Fast markdown RAG with multi-provider embeddings (Vertex AI, Gemini, OpenAI, Voyage), incremental indexing with mtime/size fast-path, stale vector pruning, Milvus vector store.
Apache-2.0 ★ 5 -
shopware-upgrade-intelligencegithub.com/vlpmedialtd/shopware-upgrade-intelligenceupdated this year
Local-first RAG-MCP server for every Shopware 6 version. Ask Claude Code about breaking changes, deprecations, CSS classes and upgrade paths across 6.4–6.7.
MIT -
Persistent multi-path memory for AI agents — vector + BM25 keyword + entity graph search with RRF fusion, credential scrubbing, auto-consolidation, multi-backend storage (Qdrant + SQLite/Postgres)
npm MIT -
@marwansaab/obsidian-modified-mcp-servergithub.com/marwansaab/obsidian-modified-mcp-serverupdated this year
Personal fork of @connorbritain/obsidian-mcp-server. TypeScript MCP server for Obsidian with wrapper-side mitigations of upstream limitations — first concrete change is re-enabling patch_content under a structural-only path validator.
npm MIT -
A governed MCP server: canonical, provenance-stamped answers for business metrics over role-based access, with a lexer-guarded ad-hoc path and a full audit trail. Architecture, decision records and a working reference implementation on synthetic data.
CC-BY-4.0 -
Local RAG MCP server for markdown docs. Retrieval only (no answer LLM): ask_docs returns grounded passages plus citations that resolve to real files on disk. Index any local path or git repo, filter by layer, embed offline.
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 -
Fully local MCP retrieval tool for coding agents. 161M encoder, ~1s per tool call, no API, no LLM in the query path. No file reads: line numbers, callers, callees and exact call sites, packed to a token budget. 2-4x the file coverage of BM25 and TF-IDF on SWE-Explore, and tops the paper's own sub-1B code model on CORE-Bench.
MIT ★ 5