A self-hosted model library. Download Hugging Face models onto your own storage, pin the revision you want, verify the copy, and keep an eye out for updates. Includes an MCP server and scoped keys for
View on GitHub ↗Homepage ↗A model library on your own storage. Find models on the Hugging Face Hub, keep exactly the files you want, and read them from a plain folder.
ModelHarbor is self-hosted software for local model weights. It is not affiliated with or endorsed by Hugging Face.
Multi-userRoles, per-account libraries, notes, and preferences. Shared notesModel notes travel with the team, pinned or latest.
Private uploadsYour own weights, private by default, shareable within the install. Storage locationsMove copies between cold, warm, and hot tiers on demand.
Revision pinningThe copy is the contract. Drift is reported, never applied. Agentic controlScoped keys and a stable API for research and ops.
One containerThe whole service is a container and a volume. Nothing else to operate.
ModelHarbor ships as a container. Running it needs Docker and two host directories: one for model files, one for application state.
Edit .env — MODELSTORAGEPATH is the host directory that holds models — then:
Open port 7860. The first visit creates the owner account, which adds member accounts from Settings. docker compose down stops the service; model files and the database are preserved.
Session cookies are marked Secure by default, so sign-ins persist once HTTPS terminates in front of the app. On a plain-HTTP LAN, the template documents the escape hatch.
Windows users can run it with Docker Desktop through the helpers in deploy/desktop; the same document covers the compose route and a source checkout in full.
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/toks-love--modelharbor.html)
Stop your AI from making things up — it proposes, deterministic tools decide, every claim checked against ground truth with evidence. Grounded facts and context survive resets. Reverse engineering is
zhuyansen/new-term-radarDaily radar for new AI terms: names of new models, products and repos, caught on X, Hugging Face, GitHub and sitemaps before keyword tools list them.
yzg1226/mcd-party-officer🍟 麦当劳门店派对成团调度 Skill — 把派对从「策划」推进到「办得成」：成团概率雷达 / 包场vs拼团 / 择日黄历 / 儿童营养指数 / 招募卡与邀请函
fellowgeek/mcp-memoryAn OKF-backed Model Context Protocol (MCP) server delivering persistent long-term memory and SQLite FTS5 search for AI agents.
hirotomasato/yowesGenerate realistic teacher documents (ID cards, licenses, letters) for 13 countries via MCP.
hujjun/mcd-saver-strategist麦当劳省钱点餐助手（麦门精算师）｜麦当劳优惠券组合优化 Skill：拉取麦当劳菜单与全部可用优惠券，算出最省的点餐组合，巨无霸套餐 ¥47 → ¥21.9（省 53%）。McDonalds MCP（麦当劳 MCP）coupon optimizer · McDonalds China MCP Server · Agent Skill for WorkBuddy · 麦当劳程序员创意开发大赛参赛作品
The top new MCP servers of the week, every Monday. No spam, unsubscribe anytime.