Hugging Face MCP server for AI agents
Let Claude, ChatGPT or your own agent read Hugging Face: 3 typed functions through one MCP server. No Hugging Face API key needed.
- Site
- huggingface.co
- Functions
- 3
- Sign-in
- Not needed
- Works with
- MCP, HTTP, Node, Python
Connect
Just Hugging Face
An MCP server for Hugging Face alone, with one tool per function. Paste it into Claude, ChatGPT, Cursor or any MCP client as a remote server.
https://api.bowmark.ai/mcp/site/huggingfaceEvery site
One MCP server for Hugging Face and every other site Bowmark covers. Your agent looks up what it needs and calls it.
https://api.bowmark.ai/mcpSet it up in your clientBoth need a Bowmark API key. The quickstart gets you one and a first result in a few minutes.
Functions
Open a function to see its arguments, what it returns and how to call it.
Read 3
getModelReturns one Hugging Face model by its repo id or huggingface.co url.Arguments
Name Type Required idstringYes Returns
HuggingFaceModelinterface HuggingFaceModel { id: string; author: string; name: string; task: string | null; library: string | null; likes: number | null; downloads: number | null; trendingScore: number | null; license: string | null; tags: string[]; createdAt: string | null; lastModified: string | null; gated: boolean | string | null; url: string; }HTTP endpoint
POST https://api.bowmark.ai/v1/sites/huggingface/getModelCall it
import { bowmark } from "@bowmark/web"; const result = await bowmark.providers.huggingface.getModel("openai-community/gpt2");import asyncio from bowmark_web import bowmark result = asyncio.run(bowmark.providers.huggingface.getModel("openai-community/gpt2"))curl -X POST https://api.bowmark.ai/v1/sites/huggingface/getModel \ -H "Authorization: Bearer $BOWMARK_API_KEY" \ -H "Content-Type: application/json" \ -d '{"id":"openai-community/gpt2"}'// The body of a run: the MCP run tool, or POST /v1/run return await bowmark.providers.huggingface.getModel("openai-community/gpt2");listTrendingModelsReturns the trending models on the Hugging Face Hub right now, in the Hub's own trending order.Arguments
Name Type Required taskstringNo limitnumberNo Returns
HuggingFaceModel[]interface HuggingFaceModel { id: string; author: string; name: string; task: string | null; library: string | null; likes: number | null; downloads: number | null; trendingScore: number | null; license: string | null; tags: string[]; createdAt: string | null; lastModified: string | null; gated: boolean | string | null; url: string; }HTTP endpoint
POST https://api.bowmark.ai/v1/sites/huggingface/listTrendingModelsCall it
import { bowmark } from "@bowmark/web"; const result = await bowmark.providers.huggingface.listTrendingModels({ limit: 3 });import asyncio from bowmark_web import bowmark result = asyncio.run(bowmark.providers.huggingface.listTrendingModels({"limit": 3}))curl -X POST https://api.bowmark.ai/v1/sites/huggingface/listTrendingModels \ -H "Authorization: Bearer $BOWMARK_API_KEY" \ -H "Content-Type: application/json" \ -d '{"args":{"limit":3}}'// The body of a run: the MCP run tool, or POST /v1/run return await bowmark.providers.huggingface.listTrendingModels({ limit: 3 });searchModelsSearches every public Hugging Face model by name, author (org) or task, sorted by trending, downloads, likes, created or updated.Arguments
Name Type Required argsstring | HuggingFaceSearchArgsYes interface HuggingFaceSearchArgs { query?: string; author?: string; task?: string; sort?: "trending" | "downloads" | "likes" | "created" | "updated"; limit?: number; }Returns
HuggingFaceModel[]interface HuggingFaceModel { id: string; author: string; name: string; task: string | null; library: string | null; likes: number | null; downloads: number | null; trendingScore: number | null; license: string | null; tags: string[]; createdAt: string | null; lastModified: string | null; gated: boolean | string | null; url: string; }HTTP endpoint
POST https://api.bowmark.ai/v1/sites/huggingface/searchModelsCall it
import { bowmark } from "@bowmark/web"; const result = await bowmark.providers.huggingface.searchModels("llama");import asyncio from bowmark_web import bowmark result = asyncio.run(bowmark.providers.huggingface.searchModels("llama"))curl -X POST https://api.bowmark.ai/v1/sites/huggingface/searchModels \ -H "Authorization: Bearer $BOWMARK_API_KEY" \ -H "Content-Type: application/json" \ -d '{"args":"llama"}'// The body of a run: the MCP run tool, or POST /v1/run return await bowmark.providers.huggingface.searchModels("llama");
Use it from code
- Node
npm i @bowmark/web - Python
pip install bowmark-web - HTTP
POST /v1/sites/huggingface/<function> - OpenAPI
/v1/sites/huggingface/openapi.json