How to integrate Hugging Face MCP with OpenAI Agents SDK

GET STARTED FOR FREE
Framework Integration Gradient
Hugging Face Logo
open-ai-agents-sdk Logo
divider

Introduction

This guide walks you through connecting Hugging Face to the OpenAI Agents SDK using the Composio tool router. By the end, you'll have a working Hugging Face agent that can run text summarization on uploaded document, list all my hugging face model repositories, deploy a new model to spaces through natural language commands.

This guide will help you understand how to give your OpenAI Agents SDK agent real control over a Hugging Face account through Composio's Hugging Face MCP server.

Before we dive in, let's take a quick look at the key ideas and tools involved.

TL;DR

Here's what you'll learn:
  • Get and set up your OpenAI and Composio API keys
  • Install the necessary dependencies
  • Initialize Composio and create a Tool Router session for Hugging Face
  • Configure an AI agent that can use Hugging Face as a tool
  • Run a live chat session where you can ask the agent to perform Hugging Face operations

What is open-ai-agents-sdk?

The OpenAI Agents SDK is a lightweight framework for building AI agents that can use tools and maintain conversation state. It provides a simple interface for creating agents with hosted MCP tool support.

Key features include:

  • Hosted MCP Tools: Connect to external services through hosted MCP endpoints
  • SQLite Sessions: Persist conversation history across interactions
  • Simple API: Clean interface with Agent, Runner, and tool configuration
  • Streaming Support: Real-time response streaming for interactive applications

What is the Hugging Face MCP server, and what's possible with it?

The Hugging Face MCP server is an implementation of the Model Context Protocol that connects your AI agent and assistants like Claude, Cursor, etc directly to your Hugging Face account. It provides structured and secure access so your agent can perform Hugging Face operations on your behalf.

Supported Tools & Triggers

Tools
Change Discussion StatusTool to change the status of a Hugging Face repository discussion.
Check Dataset ValidityTool to check whether a specific dataset is valid on Hugging Face Hub.
Check Models Upload MethodTool to check if files should be uploaded through the Large File mechanism or directly.
Check Spaces Upload MethodTool to check if files should be uploaded through the Large File mechanism or directly to Hugging Face Spaces.
Claim Paper AuthorshipTool to claim authorship of a paper on Hugging Face.
Request Repository AccessTool to request access to a gated repository on Hugging Face Hub.
Create CollectionTool to create a new collection on Hugging Face.
Create Datasets BranchTool to create a new branch in a Hugging Face dataset repository.
Create Datasets CommitTool to create a commit in a Hugging Face dataset repository.
Check Dataset File Upload MethodTool to check if files should be uploaded via Large File Storage (LFS) or directly to a Hugging Face dataset repository.
Create Datasets TagTool to create a tag on a Hugging Face dataset repository.
Create DiscussionTool to create a new discussion on a Hugging Face repository (model, dataset, or Space).
Create Discussion CommentTool to create a new comment on a Hugging Face repository discussion.
Pin discussionTool to pin or unpin a discussion on a Hugging Face repository (model, dataset, or Space).
Create models branchTool to create a new branch in a Hugging Face model repository.
Create Models CommitTool to create a commit to a Hugging Face model repository.
Create Models TagTool to create a tag on a Hugging Face model repository.
Create Paper CommentTool to create a new comment on a Hugging Face paper.
Create Papers Comment ReplyTool to create a reply to a comment on a Hugging Face paper.
Create Papers IndexTool to index a paper from arXiv by its ID on Hugging Face.
Create RepositoryTool to create a new repository (model, dataset, or Space) on Hugging Face Hub.
Create spaces branchTool to create a new branch in a Hugging Face space repository.
Create Spaces CommitTool to create a commit in a Hugging Face Space repository.
Create or update Space secretTool to create or update a secret in a Hugging Face Space.
Create Spaces TagTool to create a tag on a Hugging Face space repository.
Create or update Space variableTool to create or update a variable in a Hugging Face Space.
Create SQL Console EmbedTool to create a SQL Console embed for querying datasets on Hugging Face.
Create WebhookTool to create a webhook on Hugging Face that triggers on repository or discussion events.
Delete dataset branchTool to delete a branch from a Hugging Face dataset repository.
Delete Dataset TagTool to delete a tag from a Hugging Face dataset.
Delete discussionTool to delete a discussion from a Hugging Face repository.
Delete network CIDR listTool to delete a network CIDR list entry from Hugging Face Inference Endpoints.
Delete notificationsTool to delete notifications from Hugging Face.
Delete Settings WebhookTool to delete a webhook from Hugging Face settings.
Delete space branchTool to delete a branch from a Hugging Face space repository.
Delete space secretTool to delete a secret from a Hugging Face space.
Delete Spaces TagTool to delete a tag from a Hugging Face space.
Delete space variableTool to delete a variable from a Hugging Face space.
Filter dataset rowsTool to filter rows in a Hugging Face dataset split based on SQL-like query conditions.
Generate Chat CompletionTool to generate a response given a list of messages in a conversational context.
Generate Text EmbeddingsTool to convert text into vector embeddings for feature extraction, semantic search, and similarity tasks.
Get Daily PapersTool to retrieve daily papers from Hugging Face.
Get Dataset Croissant MetadataTool to get Croissant metadata about a Hugging Face dataset.
Get Dataset First RowsTool to get the first 100 rows of a dataset split along with column data types and features.
Get Dataset InfoTool to get general information about a dataset including description, citation, homepage, license, and features (column schemas).
Get Dataset Repository InfoTool to retrieve detailed information about a Hugging Face dataset repository.
Get Dataset RowsTool to retrieve a slice of rows from a Hugging Face dataset split at any given location (offset).
Get Datasets CompareTool to get a comparison (diff) between two revisions of a Hugging Face dataset.
Get Dataset SizeTool to get the size of a Hugging Face dataset including number of rows and size in bytes.
Get Datasets JWTTool to generate a JWT token for accessing a Hugging Face dataset repository.
Get Datasets LeaderboardTool to retrieve evaluation results ranked by score for a dataset's leaderboard.
Get Dataset Notebook URLTool to get a Jupyter notebook URL from a Hugging Face dataset repository.
Get Datasets ResolveTool to resolve and download a file from a Hugging Face dataset repository.
Get Dataset Security ScanTool to retrieve the security scan status of a Hugging Face dataset repository.
Get Dataset Tags by TypeTool to retrieve all possible tags used for datasets on Hugging Face, grouped by tag type.
Get Dataset StatisticsTool to get comprehensive statistics about a dataset split including column statistics and data distribution information.
Get dataset repository sizeTool to get the total size of a Hugging Face dataset repository at a specific revision and path.
Get Datasets XET Read TokenTool to get a read short-lived access token for XET from Hugging Face datasets.
Get Discussion DetailsTool to get detailed information about a specific discussion or pull request on Hugging Face Hub.
Get Discussion DetailsTool to retrieve discussion details from a Hugging Face repository.
Get Available Job HardwareTool to retrieve available hardware configurations for Hugging Face Jobs with their specifications and pricing.
Get Model InformationTool to retrieve detailed information about a Hugging Face model repository.
Get Models CompareTool to compare two revisions of a Hugging Face model repository.
Get Models JWTTool to generate a JWT token for accessing a Hugging Face model repository.
Get Models NotebookTool to retrieve a Jupyter notebook URL from a Hugging Face model repository.
Get Model Security Scan StatusTool to retrieve the security scan status of a Hugging Face model repository.
Get model repository sizeTool to get the total size of a Hugging Face model repository at a specific revision and path.
Get Model XET Read TokenTool to retrieve a short-lived XET read access token for a Hugging Face model repository.
Get Model Tags By TypeTool to retrieve all possible tags used for Hugging Face models, grouped by tag type.
Get Organization AvatarTool to retrieve the avatar URL for a Hugging Face organization.
Get Organization MembersTool to retrieve a list of members for a Hugging Face organization.
Get Organization Social HandlesTool to retrieve an organization's social media handles from Hugging Face.
Get ResolveTool to resolve a file in a Hugging Face repository.
Get Resolve Cache DatasetsTool to resolve a file from cache in a Hugging Face dataset repository.
Get Resolve Cache ModelsTool to resolve and retrieve files from the Hugging Face model cache.
Get Resolve Cache SpacesTool to resolve and retrieve a file from Hugging Face Spaces cache.
Get Billing UsageTool to retrieve user billing usage for a given period from Hugging Face.
Get Jobs UsageTool to retrieve Jobs usage and billing information for the current subscription period from Hugging Face.
Get Live Billing UsageTool to retrieve live billing usage stream from Hugging Face.
Get Billing Usage V2Tool to retrieve user billing usage for a custom date range from Hugging Face.
Get MCP SettingsTool to retrieve MCP (Model Context Protocol) tools configuration for the authenticated user.
Get Settings WebhookTool to retrieve a specific webhook configuration from Hugging Face settings.
Get Space InfoTool to retrieve detailed information about a Hugging Face Space repository.
Get Spaces CompareTool to compare two revisions of a Hugging Face Space repository.
Get Spaces EventsTool to stream status updates for a Hugging Face Space using SSE protocol.
Get Spaces JWTTool to generate a JWT token for accessing a Hugging Face space repository.
Get Space MetricsTool to get live metrics for a specific Space in a streaming fashion, with SSE protocol, such as current Zero-GPU usage.
Get Spaces NotebookTool to retrieve a Jupyter notebook URL from a Hugging Face space repository.
Get Spaces ResolveTool to resolve and retrieve a file from a Hugging Face Space repository.
Get Space Security Scan StatusTool to retrieve the security scan status of a Hugging Face space repository.
Get space repository sizeTool to get the total size of a Hugging Face space repository at a specific revision and path.
Get Space XET Read TokenTool to retrieve a short-lived XET read access token for a Hugging Face Space repository.
Get Spaces XET Write TokenTool to retrieve a short-lived XET write access token for a Hugging Face space repository.
Get Trending RepositoriesTool to retrieve trending repositories from Hugging Face.
Get User AvatarTool to retrieve the avatar URL for a Hugging Face user.
Get User OverviewTool to retrieve a comprehensive overview of a Hugging Face user's profile.
Get User Social HandlesTool to retrieve a user's social media handles from Hugging Face.
Get Authenticated User InfoTool to get information about the authenticated Hugging Face user including username, email, organizations, and token details.
Handle Dataset User Access RequestTool to handle a user's access request to a gated Hugging Face dataset.
List CollectionsTool to list collections on the Hugging Face Hub.
List Dataset Parquet FilesTool to get the list of Parquet files for a dataset.
List Dataset Paths InfoTool to list detailed information about specific paths in a Hugging Face dataset repository.
List DatasetsTool to list datasets on the Hugging Face Hub.
List Dataset CommitsTool to list commits from a Hugging Face dataset repository.
List Dataset SplitsTool to get the list of subsets and splits of a dataset.
List Dataset ReferencesTool to list all references (branches, tags, converts, pull requests) in a Hugging Face dataset repository.
List datasets treeTool to list the content of a Hugging Face dataset repository tree with pagination support.
List Dataset Access RequestsTool to list access requests for a gated Hugging Face dataset repository.
List DiscussionsTool to list discussions for a Hugging Face repository.
List Repository DiscussionsTool to list discussions and pull requests for a Hugging Face repository.
List Available DocumentationTool to retrieve the list of available documentation from Hugging Face.
List Inference EndpointsTool to list Hugging Face Inference Endpoints for a specific user or organization.
List models on Hugging Face HubTool to list models on the Hugging Face Hub with filtering options.
List Model CommitsTool to list commits from a Hugging Face model repository.
List Model Paths InfoTool to list detailed information about specific paths in a Hugging Face model repository.
List Model ReferencesTool to list all references (branches, tags, converts, and optionally pull requests) in a Hugging Face model repository.
List model repository treeTool to list the contents of a Hugging Face model repository tree at a specific revision and path, with pagination support.
List NotificationsTool to list notifications for the authenticated Hugging Face user.
List Repository CommitsTool to list commits for a Hugging Face repository.
List repository filesTool to get the file tree of a Hugging Face repository with pagination support.
List WebhooksTool to list all webhooks configured in Hugging Face settings.
List Spaces on Hugging Face HubTool to list Spaces on the Hugging Face Hub with filtering options.
List Space CommitsTool to list commits from a Hugging Face Space repository.
List Available Space HardwareTool to retrieve available hardware configurations for Hugging Face Spaces with their specifications and pricing.
List Space LFS FilesTool to list LFS (Large File Storage) files from a Hugging Face Space repository.
List Space Paths InfoTool to list detailed information about specific paths in a Hugging Face space repository.
List Space ReferencesTool to list all references (branches, tags, converts, pull requests) in a Hugging Face space repository.
List spaces treeTool to list the content of a Hugging Face space repository tree with pagination support.
List Cloud Provider VendorsTool to list available cloud provider vendors for Hugging Face Inference Endpoints.
Search datasetTool to search text in a dataset split on Hugging Face.
Search DocumentationTool to search Hugging Face documentation across all products and libraries.
Search PapersTool to perform hybrid semantic/full-text search on papers in Hugging Face.
Squash Dataset CommitsTool to squash all commits in a dataset ref into a single commit with the given message.
Squash Spaces CommitsTool to squash all commits in a space ref into a single commit with the given message.
Update Dataset SettingsTool to update settings for a Hugging Face dataset repository.
Update Discussion TitleTool to change the title of an existing discussion on a Hugging Face repository (model, dataset, or Space).
Update Model Repository SettingsTool to update settings for a Hugging Face model repository.
Update Notification SettingsTool to update notification settings for the authenticated Hugging Face user.
Update Watch SettingsTool to update watch settings for your Hugging Face account.
Update WebhookTool to update an existing webhook in Hugging Face settings.
Update Spaces Repository SettingsTool to update settings for a Hugging Face Spaces repository.
Update SQL Console EmbedTool to update an existing SQL console embed for a Hugging Face dataset.
Enable or Disable WebhookTool to enable or disable a webhook on Hugging Face.

What is the Composio tool router, and how does it fit here?

What is Tool Router?

Composio's Tool Router helps agents find the right tools for a task at runtime. You can plug in multiple toolkits (like Gmail, HubSpot, and GitHub), and the agent will identify the relevant app and action to complete multi-step workflows. This can reduce token usage and improve the reliability of tool calls. Read more here: Getting started with Tool Router

The tool router generates a secure MCP URL that your agents can access to perform actions.

How the Tool Router works

The Tool Router follows a three-phase workflow:

  1. Discovery: Searches for tools matching your task and returns relevant toolkits with their details.
  2. Authentication: Checks for active connections. If missing, creates an auth config and returns a connection URL via Auth Link.
  3. Execution: Executes the action using the authenticated connection.

Step-by-step Guide

Prerequisites

Before starting, make sure you have:
  • Composio API Key and OpenAI API Key
  • Primary know-how of OpenAI Agents SDK
  • A live Hugging Face project
  • Some knowledge of Python or Typescript

Getting API Keys for OpenAI and Composio

OpenAI API Key
  • Go to the OpenAI dashboard and create an API key. You'll need credits to use the models, or you can connect to another model provider.
  • Keep the API key safe.
Composio API Key

Install dependencies

pip install composio_openai_agents openai-agents python-dotenv

Install the Composio SDK and the OpenAI Agents SDK.

Set up environment variables

bash
OPENAI_API_KEY=sk-...your-api-key
COMPOSIO_API_KEY=your-api-key
USER_ID=composio_user@gmail.com

Create a .env file and add your OpenAI and Composio API keys.

Import dependencies

import asyncio
import os
from dotenv import load_dotenv

from composio import Composio
from composio_openai_agents import OpenAIAgentsProvider
from agents import Agent, Runner, HostedMCPTool, SQLiteSession
What's happening:
  • You're importing all necessary libraries.
  • The Composio and OpenAIAgentsProvider classes are imported to connect your OpenAI agent to Composio tools like Hugging Face.

Set up the Composio instance

load_dotenv()

api_key = os.getenv("COMPOSIO_API_KEY")
user_id = os.getenv("USER_ID")

if not api_key:
    raise RuntimeError("COMPOSIO_API_KEY is not set. Create a .env file with COMPOSIO_API_KEY=your_key")

# Initialize Composio
composio = Composio(api_key=api_key, provider=OpenAIAgentsProvider())
What's happening:
  • load_dotenv() loads your .env file so OPENAI_API_KEY and COMPOSIO_API_KEY are available as environment variables.
  • Creating a Composio instance using the API Key and OpenAIAgentsProvider class.

Create a Tool Router session

# Create a Hugging Face Tool Router session
session = composio.create(
    user_id=user_id,
    toolkits=["hugging_face"]
)

mcp_url = session.mcp.url

What is happening:

  • You give the Tool Router the user id and the toolkits you want available. Here, it is only hugging_face.
  • The router checks the user's Hugging Face connection and prepares the MCP endpoint.
  • The returned session.mcp.url is the MCP URL that your agent will use to access Hugging Face.
  • This approach keeps things lightweight and lets the agent request Hugging Face tools only when needed during the conversation.

Configure the agent

# Configure agent with MCP tool
agent = Agent(
    name="Assistant",
    model="gpt-5",
    instructions=(
        "You are a helpful assistant that can access Hugging Face. "
        "Help users perform Hugging Face operations through natural language."
    ),
    tools=[
        HostedMCPTool(
            tool_config={
                "type": "mcp",
                "server_label": "tool_router",
                "server_url": mcp_url,
                "headers": {"x-api-key": api_key},
                "require_approval": "never",
            }
        )
    ],
)
What's happening:
  • We're creating an Agent instance with a name, model (gpt-5), and clear instructions about its purpose.
  • The agent's instructions tell it that it can access Hugging Face and help with queries, inserts, updates, authentication, and fetching database information.
  • The tools array includes a HostedMCPTool that connects to the MCP server URL we created earlier.
  • The headers dict includes the Composio API key for secure authentication with the MCP server.
  • require_approval: 'never' means the agent can execute Hugging Face operations without asking for permission each time, making interactions smoother.

Start chat loop and handle conversation

print("\nComposio Tool Router session created.")

chat_session = SQLiteSession("conversation_openai_toolrouter")

print("\nChat started. Type your requests below.")
print("Commands: 'exit', 'quit', or 'q' to end\n")

async def main():
    try:
        result = await Runner.run(
            agent,
            "What can you help me with?",
            session=chat_session
        )
        print(f"Assistant: {result.final_output}\n")
    except Exception as e:
        print(f"Error: {e}\n")

    while True:
        user_input = input("You: ").strip()
        if user_input.lower() in {"exit", "quit", "q"}:
            print("Goodbye!")
            break

        result = await Runner.run(
            agent,
            user_input,
            session=chat_session
        )
        print(f"Assistant: {result.final_output}\n")

asyncio.run(main())
What's happening:
  • The program prints a session URL that you visit to authorize Hugging Face.
  • After authorization, the chat begins.
  • Each message you type is processed by the agent using Runner.run().
  • The responses are printed to the console, and conversations are saved locally using SQLite.
  • Typing exit, quit, or q cleanly ends the chat.

Complete Code

Here's the complete code to get you started with Hugging Face and open-ai-agents-sdk:

import asyncio
import os
from dotenv import load_dotenv

from composio import Composio
from composio_openai_agents import OpenAIAgentsProvider
from agents import Agent, Runner, HostedMCPTool, SQLiteSession

load_dotenv()

api_key = os.getenv("COMPOSIO_API_KEY")
user_id = os.getenv("USER_ID")

if not api_key:
    raise RuntimeError("COMPOSIO_API_KEY is not set. Create a .env file with COMPOSIO_API_KEY=your_key")

# Initialize Composio
composio = Composio(api_key=api_key, provider=OpenAIAgentsProvider())

# Create Tool Router session
session = composio.create(
    user_id=user_id,
    toolkits=["hugging_face"]
)
mcp_url = session.mcp.url

# Configure agent with MCP tool
agent = Agent(
    name="Assistant",
    model="gpt-5",
    instructions=(
        "You are a helpful assistant that can access Hugging Face. "
        "Help users perform Hugging Face operations through natural language."
    ),
    tools=[
        HostedMCPTool(
            tool_config={
                "type": "mcp",
                "server_label": "tool_router",
                "server_url": mcp_url,
                "headers": {"x-api-key": api_key},
                "require_approval": "never",
            }
        )
    ],
)

print("\nComposio Tool Router session created.")

chat_session = SQLiteSession("conversation_openai_toolrouter")

print("\nChat started. Type your requests below.")
print("Commands: 'exit', 'quit', or 'q' to end\n")

async def main():
    try:
        result = await Runner.run(
            agent,
            "What can you help me with?",
            session=chat_session
        )
        print(f"Assistant: {result.final_output}\n")
    except Exception as e:
        print(f"Error: {e}\n")

    while True:
        user_input = input("You: ").strip()
        if user_input.lower() in {"exit", "quit", "q"}:
            print("Goodbye!")
            break

        result = await Runner.run(
            agent,
            user_input,
            session=chat_session
        )
        print(f"Assistant: {result.final_output}\n")

asyncio.run(main())

Conclusion

This was a starter code for integrating Hugging Face MCP with OpenAI Agents SDK to build a functional AI agent that can interact with Hugging Face.

Key features:

  • Hosted MCP tool integration through Composio's Tool Router
  • SQLite session persistence for conversation history
  • Simple async chat loop for interactive testing
You can extend this by adding more toolkits, implementing custom business logic, or building a web interface around the agent.

How to build Hugging Face MCP Agent with another framework

Claude Agent SDK

Claude Agent SDK

Use Hugging Face MCP with Claude Agent SDK

Google ADK

Google ADK

Use Hugging Face MCP with Google ADK

LangChain

LangChain

Use Hugging Face MCP with LangChain

Vercel AI SDK

Vercel AI SDK

Use Hugging Face MCP with Vercel AI SDK

Mastra AI

Mastra AI

Use Hugging Face MCP with Mastra AI

LlamaIndex

LlamaIndex

Use Hugging Face MCP with LlamaIndex

CrewAI

CrewAI

Use Hugging Face MCP with CrewAI

Claude Code

Claude Code

Use Hugging Face MCP with Claude Code

Codex

Codex

Use Hugging Face MCP with Codex

FAQ

What are the differences in Tool Router MCP and Hugging Face MCP?

With a standalone Hugging Face MCP server, the agents and LLMs can only access a fixed set of Hugging Face tools tied to that server. However, with the Composio Tool Router, agents can dynamically load tools from Hugging Face and many other apps based on the task at hand, all through a single MCP endpoint.

Can I use Tool Router MCP with OpenAI Agents SDK?

Yes, you can. OpenAI Agents SDK fully supports MCP integration. You get structured tool calling, message history handling, and model orchestration while Tool Router takes care of discovering and serving the right Hugging Face tools.

Can I manage the permissions and scopes for Hugging Face while using Tool Router?

Yes, absolutely. You can configure which Hugging Face scopes and actions are allowed when connecting your account to Composio. You can also bring your own OAuth credentials or API configuration so you keep full control over what the agent can do.

How safe is my data with Composio Tool Router?

All sensitive data such as tokens, keys, and configuration is fully encrypted at rest and in transit. Composio is SOC 2 Type 2 compliant and follows strict security practices so your Hugging Face data and credentials are handled as safely as possible.

Used by agents from

Context
Letta
glean
HubSpot
Agent.ai
Altera
DataStax
Entelligence
Rolai
Context
Letta
glean
HubSpot
Agent.ai
Altera
DataStax
Entelligence
Rolai
Context
Letta
glean
HubSpot
Agent.ai
Altera
DataStax
Entelligence
Rolai

Never worry about agent reliability

We handle tool reliability, observability, and security so you never have to second-guess an agent action.

Book a call↗Get started for free