Hugging Face MCP connector for AI agents

Tools your AI agent reaches for on Hugging Face MCP, scoped per user.

CALL ANY TOOL

Hugging Face MCP tools give agents access to Hub search across Spaces, papers, and repositories, detailed repo metadata retrieval, authenticated user profile lookup, natural language Hub queries, paginated documentation fetching, and AI image generation.

huggingfacemcp_dynamic_space

Dynamic Space

Call a Hugging Face MCP-enabled Space dynamically. Use 'discover' to list available MCP spaces, 'view_parameters' to inspect a space's tools, or 'invoke' to call a specific tool.

Parameters

Name

Type

Required

Description

operation

string

Optional

Action to perform: discover lists available tasks, view_parameters inspects a space, invoke runs it.

parameters

string

Optional

JSON string of input parameters. Used only with the "invoke" operation.

space_name

string

Optional

Space ID in "username/space-name" format. Required for "view_parameters" and "invoke" operations.

huggingfacemcp_gr1_z_image_turbo_generate

Gr1 Z Image Turbo Generate

huggingfacemcp_hf_doc_fetch

Fetch hf doc

huggingfacemcp_hf_doc_search

Search hf doc

huggingfacemcp_hf_hub_query

Query hf hub

huggingfacemcp_hf_whoami

Hf Whoami

huggingfacemcp_hub_repo_details

Hub Repo Details

huggingfacemcp_hub_repo_search

Search hub repo

huggingfacemcp_paper_search

Search paper

huggingfacemcp_space_search

Search space

Build your Agent

Same auth pattern across every framework.

from langchain_mcp_adapters.client import MultiServerMCPClient
from scalekit import ScalekitClient

client = ScalekitClient(env_url=ENV_URL, client_id=CLIENT_ID, client_secret=SECRET)
token = client.agent.get_token(user_id="user_id", connector="huggingfacemcp")

mcp = MultiServerMCPClient({
    "huggingfacemcp": {
        "url": "https://mcp.scalekit.com/huggingfacemcp",
        "headers": {"Authorization": "Bearer " + token}
    }
})
tools = await mcp.get_tools()

import OpenAI from "openai";
import { ScalekitClient } from "@scalekit-sdk/node";

const client = new ScalekitClient({ envUrl, clientId, clientSecret });
const token = await client.agent.getToken({ userId: "user_id", connector: "huggingfacemcp" });

const openai = new OpenAI();
// Connect to MCP at https://mcp.scalekit.com/huggingfacemcp
// Pass: Authorization: Bearer + token

import Anthropic from "@anthropic-ai/sdk";
import { ScalekitClient } from "@scalekit-sdk/node";

const client = new ScalekitClient({ envUrl, clientId, clientSecret });
const token = await client.agent.getToken({ userId: "user_id", connector: "huggingfacemcp" });

const anthropic = new Anthropic();
// Connect to MCP at https://mcp.scalekit.com/huggingfacemcp
// Pass: Authorization: Bearer + token

from google.adk.agents import LlmAgent
from scalekit import ScalekitClient

client = ScalekitClient(env_url=ENV_URL, client_id=CLIENT_ID, client_secret=SECRET)
token = client.agent.get_token(user_id="user_id", connector="huggingfacemcp")
# Connect to MCP at https://mcp.scalekit.com/huggingfacemcp
# Pass: Authorization: Bearer + token

Try these prompts

Paste any prompt into your agent to get started.

Search & discover

Copy the prompt

Copied

Search for the top 10 most-liked image generation Spaces on Hugging Face and summarize what each one does.

Copy the prompt

Copied

Find all papers on Hugging Face related to retrieval-augmented generation published in the last 6 months.

Copy the prompt

Copied

Look up the repository details for 'mistralai/Mistral-7B-v0.1' and tell me its license, tags, and download count.

Copy the prompt

Copied

Search Hugging Face for open-source speech-to-text models fine-tuned on multilingual data.

Generate & explore

Copy the prompt

Copied

Generate an image of a futuristic city skyline at sunset using the Hugging Face Image Turbo model.

Copy the prompt

Copied

Fetch the Hugging Face documentation page for the Transformers library and summarize the quickstart guide.

Copy the prompt

Copied

Ask the Hugging Face Hub what the most downloaded model for text classification is right now.

Copy the prompt

Copied

Who am I logged in as on Hugging Face? Show my profile details.

SEE HOW AUTH WORKS

User authorises once. Every agent call after uses their token with scope enforcement.

Authorize

Your user connects

Hugging Face MCP

once. We tie it to their identity and the meetings they approved — no shared bot account, no org-wide access

Who:

user ‘A’

when:

Once per user

access:

Limited to user

Store

Their

Hugging Face MCP

token lives in a vault scoped to them. User A's meetings are never reachable by an agent acting for user B, even on the same connection

vault:

encrypted

scope:

per-user

tokens:

auto-refreshed

Resolve

When your agent calls a

Hugging Face MCP

tool, we fetch the right token server-side. It never touches your agent, never appears in the LLM context, never shows up in your logs

speed:

~40ms

check:

before every call

seen by:

nobody

Audit

Every

Hugging Face MCP

tool call is logged — who triggered it, which meeting was fetched, what came back. 90 days of history, tied to the user who authorized it

history:

90 days

export:

SIEM-ready

logged:

every call

Test other agents

See the same per-user auth pattern across other connectors.

SUPPORT

Freshdesk CSAT follow-up agent

Google ADK agent that watches low CSAT scores in Freshdesk and drafts personalised follow-ups for support leads.

Freshdesk

View agent

OPS

Email-to-calendar scheduling agent

Parse scheduling intent from Gmail threads and create Google Calendar events with the right attendees and timezone.

Gmail

Google Calendar

View agent

SUPPORT

Support ticket automation (Google ADK)

Google ADK agent that classifies Zendesk tickets, pulls Notion context, and posts to Slack. End-to-end ticket handoff.

Why Scalekit

Secure your agent's access. Connectors ship in minutes

Other connector libraries treat auth as a demo afterthought.

01.

Shared tokens break per-user analytics

A shared token looks fine in a demo. In production every call looks like a service account. Scalekit resolves the real user credential.

// shared token
 audit → bot_service_account

 // scalekit
 audit → user_abc ✓

02.

Authentication is not authorization

“Our agents act across Salesforce, Gong, Google Drive, and more, on behalf of every customer. Scalekit behind the scenes meant we can keep adding tools without ever rebuilding how credentials or tool calling work.”

Venu Madhav Kattagoni

Head of Engineering / Von

FAQs

Frequently Asked Questions

Does the agent access Hugging Face MCP as the user or as a shared key?

Scalekit resolves a per-user credential for each call. The Hugging Face MCP audit log shows the real user identity, not a shared service account.

Where is the Hugging Face MCP token stored?

In Scalekit's AES-256 encrypted vault, isolated per tenant. Tokens are auto-refreshed before expiry and revocable from a single dashboard without code changes.

Can I limit what the agent is allowed to do in Hugging Face MCP?

Pass a toolNames filter to listScopedTools. Scalekit enforces the allowed set before every API call — the agent cannot invoke tools outside the approved list.

What happens when a user revokes Hugging Face MCP access?

The token is invalidated at the next tool call. The agent fails closed rather than falling back to a shared key. Other users are unaffected and the revocation event is logged with a timestamp.

Hugging Face API keys have rate limits tied to the account tier. How does Scalekit handle cases where a user's quota is exhausted mid-session?

Scalekit surfaces the upstream rate-limit error directly to the calling agent rather than silently retrying under a different credential. Your agent code receives the error and can prompt the user to review their Hugging Face plan or wait for quota reset. No shared fallback key is used, so exhaustion for one user never consumes another tenant's quota.

On this page

Tool and actions Framework Compatibility Starter Prompt Auth Lifecycle Try Other Agents Why Scalekit