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.
-
FastMCP 3.2 server plus Webapp for using the leaked Claude Code locally with ollama and open weight models. Kairos and cat included
MIT ★ 7 -
Local, offline web studio for game, video, podcast and social assets: images, 3D models, SFX, music and voices from ~100 open-weight models. Shared asset library, single-GPU job queue, prompt improver, commercial-licence filter, loudness-matched export presets and a built-in MCP server.
Apache-2.0 -
A local-first, open-weights production suite for ComfyUI. Your agent runs image, video, speech, and music workflows on your own GPU.
MIT ★ 9 -
MCP for Blender Using Open Weights Models
pypi MIT ★ 123 -
Open-weight video generation (Wan 2.2, MiniMax-H3) as MCP tools on Amazon Bedrock AgentCore Gateway, with a scale-to-zero spot GPU worker. One cdk deploy.
Apache-2.0 -
Open-source AI phone receptionist: answers in Hindi/Gujarati/English, books real appointments, exposes them to Alan agents over MCP. Open-weight models, MIT.
MIT -
Open-source AI video studio on your own ComfyUI: many open-weights workflows, a timeline editor, and MCP for Claude/ChatGPT
MIT ★ 4 -
Local AI inference infrastructure for East Africa — Ollama, open weights, degraded-mode fallbacks. 6 tools. Thesis L4.
MIT -
MCP server for quelllm.fr — 250+ open-weights LLM catalog. Tools: list/compare/estimate VRAM/estimate cost API vs self-hosted.
MIT -
ZeroFusegithub.com/junainfinity/ZeroFuseupdated this year
Automated, capability-preserving abliteration for open-weight LLMs — agent-native (MCP server). Clean-room MIT implementation.
MIT ★ 8 -
Open-source icon library on a single 24×24 grid. Four weights, tree-shakeable React components, a Figma plugin and an MCP server for AI agents. MIT.
npm MIT ★ 1 -
Borgo: compare any number of cities on open data, weighted by what you value: take-home pay, money left after rent, right to live, climate. Web app, CLI and MCP server.
npm Apache-2.0 -
MCP server for AbliteratedLLM: list models and prices, estimate cost, check balance, and chat with abliterated and stock open-weight models.
npm MIT -
A lazy prompt fixes itself on open-weight models (Gemma 3, gpt-oss) over AcruxCore's MCP server. Banking77, held-out exact-match test.
Apache-2.0 -
prog-strength-mcpgithub.com/Prog-Strength/prog-strength-mcpupdated this year
Lets AI agents reason over your real lifting data instead of guessing. The MCP server for Prog Strength, an open-source AI-native weight training tracker.
-
IT/customer-support agent: MCP tool server + LangGraph, human approval for risky actions, prompt-injection & PII-leak evals. Local open-weight models (Ollama).
-
Remote MCP server and agent skill for Local AI Picks: open-weight models that fit your hardware
MIT -
MCP servers that let LLM agents operate Juju safely: server-side policy guardrails, audit log, Prometheus-driven autoscaling, and an eval harness for open-weight models.
Apache-2.0 -
lechatai-mcpgithub.com/rocnubie/lechatai-mcpupdated this year
LeChat AI (lechatai.online) is an independent guide and resource hub for Mistral's Le Chat conversational AI platform. The site covers the platform's capabilities in depth — from its speed benchmarks and pricing tiers to its European data compliance stance and open-weight model philosophy.
MIT -
JevStyle: local decision models for AI agents. Classify, route and score with a probability per option. systemone-compatible local server, MCP tools, Claude Code guard. Open weights, Apache-2.0.
pypi Apache-2.0 ★ 135 -
Open-source platform for AI chat agents that answer from your own knowledge base. Hybrid RAG search in PostgreSQL, MCP and custom tools, guardrails and human handoff. Put agents on a hosted chat page, a website widget, Slack, Teams, WhatsApp, Telegram or a REST API. Self-host with Docker and open-weight models.
NOASSERTION ★ 54 -
Mstari - the Triage Desk: an agentic triage queue for public outpatient departments. Cited SATS priority scoring, own MCP server (6 tools, 2 gated), LangGraph, local open-weights Qwen, named-clinician human gate. Agentic AI for Africa Innovation Challenge - healthcare track.
MIT -
A self-hosted, containerized platform for AI agents, exposed as Capability Packs — schema-validated, one-shot JSON tools — and native MCP. The defining metric is ≥90% pack success on 7B–30B-class open-weight models, something no frontier-targeting competitor is optimizing for.
npm Apache-2.0 ★ 6 -
polymath-megablaster-mcpactive
Deterministic calculation toolkit for AI agents, especially local/open-weight models (Llama, Qwen, Mistral, etc.) that reliably fail at exact math, dates, and combinatorics: exact arithmetic, date math, unit conversion, statistics, subnet math, and more.
npm MIT -
rag-pipeline-grounded-llm-genomicsgithub.com/cgjuan01/rag-pipeline-grounded-llm-genomicsupdated this year
LLM · RAG · MCP: a retrieval-augmented pipeline that answers questions over my ageing & exercise genomics research from papers and gene-level results, with cited, grounded answers, plus an MCP server exposing retrieval, citation-checking and refusal as agent tools. Open-weight & API backends.
-
Tool using agent and physics simulator that recommends thermal drift mitigation strategies for deep space photonic instruments across material and environment combinations. Runs on AWS Bedrock or a self-hosted open-weight model, grounded in a 40K scenario knowledge store and an XGBoost classifier.
Apache-2.0 ★ 2 -
AI transformation infrastructure. A self-hosted knowledge base, agent board and MCP connection for putting AI to work inside a business. Run it in-house, license it as a consultant, or have us set up your server and run it for you. Works with Claude, Codex, Grok or open-weight models. Apache-2.0.
Apache-2.0 ★ 8 -
Any interface, compiled for a small model. Meta's frontier models learn an app once; a small open model runs it on your computer, every step checked, as a tool any agent can call over MCP.
Apache-2.0 -
linkedin-mcp-analyzergithub.com/outpost2026/linkedin-mcp-analyzerupdated this year
Agentic MCP pipeline that transforms LinkedIn saved jobs into structured career intelligence. Deterministic EROI scoring engine with skill gap analysis, golden-rule weighting, and automated KB write-back. Fully customizable — plug in your own methodology.