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.
-
Train PyTorch on every GPU you can reach: one Trainer, from your laptop to many machines. NVIDIA CUDA, Apple Metal and CPU; self-hosted, open source.
pypi Apache-2.0 -
MCP server and MLOps CLI to export, quantize, audit, benchmark and auto-tune PyTorch and ONNX computer-vision models for edge deployment.
MIT -
Local AI BGM and sound effects for Claude Code and any MCP client. Stable Audio 3 on Apple Silicon (MLX) and Windows NVIDIA GPUs (PyTorch). No cloud, no API key.
MIT ★ 670 -
An MCP server exposing local embedding tools - semantic search, comparison, clustering and zero-shot classification - so an agent can call them in a loop without an API round trip. Caching the embeddings took a search from 1040ms to 1.2ms. PyTorch, sentence-transformers, MCP Python SDK v2.
MIT -
velociraggithub.com/haseebkhalid1507/velociragupdated this year
Lightning-fast RAG for AI agents. ONNX-powered, 4-layer fusion, MCP server. No PyTorch.
pypi MIT -
MCP server wrapping Kyutai Pocket TTS for Claude Code (ONNX Runtime, no PyTorch, streaming, gap-free playback, French + English)
pypi MIT -
VelociRAGgithub.com/HaseebKhalid1507/VelociRAGupdated this year
Lightning-fast RAG for AI agents. ONNX-powered, 4-layer fusion, MCP server. No PyTorch.
MIT ★ 12 -
Schematic capture for neural network architectures. Draw the model, get the numbers, generate the PyTorch.
NOASSERTION -
AI-native quantum computing SDK — PyTorch QuantumLayer & Continuous Resonance, Apple Silicon Metal/MLX GPU engine, 23 MCP tools, Willow dynamic QEC, 200+ qubit MPS
pypi Apache-2.0 -
Static security scanner for machine-learning model files. Checks PyTorch, pickle, joblib, GGUF, SafeTensors, ONNX and Keras checkpoints for code execution before you load them. CLI, GitHub Action, load guard, MCP server.
MIT ★ 2 -
LynxMCP: local MCP server for the code questions grep can't answer. Call graph and blast radius, hybrid semantic + lexical code search on ONNX Runtime, library docs and PDFs as sources. No cloud, no PyTorch.
Apache-2.0 ★ 11 -
LynxMCP: 100% local MCP server for the code questions grep can't answer. Call graph and blast radius, hybrid BM25 + dense code search on ONNX Runtime, library docs and PDFs as sources. No cloud, no PyTorch.
pypi Apache Software License -
—
MIT ★ 23 -
MCP server + REST API para validacao de CPFs e consulta ao TRT3 com solver de CAPTCHA CRNN
MIT ★ 22 -
correios-rastreamentogithub.com/opastorello/correios-rastreamentoupdated this year
MCP server + REST API para rastreamento de objetos dos Correios com solver de CAPTCHA CRNN (99.6% acuracia)
MIT ★ 2 -
Scientific Paper Reverse Engineering & Replication: Local-first CLI + MCP server for literature gap discovery and mathematically verified code replicas.
-
MCP server exposing a Neurarch model graph to Claude Code, Cursor, Windsurf, and other MCP-aware AI agents.
npm MIT ★ 1 -
Denpex MCP Server: Diagnose GPU & ML training failures (CUDA OOM, NCCL timeouts, NVIDIA Xid, Slurm, vLLM) from Claude Desktop, Cursor, Windsurf, and Cline.
-
Point it at a slow training run: it finds the bottleneck, tries the fixes, and keeps only speedups it can prove with the loss unchanged.
MIT -
AI-Agent-First Continuum Physical Compiler, Universal Simulator Enhancer (.xml/.usda/.urdf/.gltf) & 200Hz Tactile Safety Shield (RT-Patch-IPC)
NOASSERTION ★ 2 -
🎙️ 免费好用的本地 AI 人声伴奏分离神器!基于 Meta Demucs 模型,一键离线无损提取 人声/伴奏/鼓点/贝斯,不限音频时长、零排队、100% 保护隐私,免装 Python 开箱即用,自带桌面端与 MCP 接口!
MIT -
Production diagnostics, analytical VRAM budgeting, ComfyUI/CLI compiler, and Model Context Protocol (FastMCP) bridge for Wan-Video (Wan2.1 & Wan2.2 MoE).
Apache-2.0 -
MOSS-SoundEffect_v2.0_MPS_ROCmgithub.com/VladimirTalyzin/MOSS-SoundEffect_v2.0_MPS_ROCmupdated this year
Desktop app for the MOSS-SoundEffect-v2.0 text-to-audio diffusion model, with working AMD ROCm (Windows) and Apple Silicon (MPS) support. No API keys, no browser. Includes CLI, benchmark tools and an MCP server.
Apache-2.0 -
Graph-Augmented Code Intelligence RAG engine pairing deterministic AST call graphs with dense vector search to boost direct-callee Recall@8 from 0.0% to 53.3% (+53.3 pp) under a 5.8ms p95 latency bound. Powered by a QLoRA fine-tuned Qwen2.5 adapter (70.02 CodeBLEU), quality verification harness, and production Model Context Protocol (MCP) server.
MIT