Which AI clients can use the Hugging Face MCP server?

27 AI clients, including Cursor, VS Code, and Claude. Each connects to Hugging Face as a remote MCP server through Runlayer, which adds identity, policy, scanning, and audit to every call.

Which Hugging Face tools are available?

Hugging Face reports its tools when your organization connects it. Admins then choose which ones each group can call.

How does Hugging Face authenticate?

Sign in the way Hugging Face requires. Runlayer holds the credentials so they stay out of the client’s configuration.

Integrations / MCP servers

Hugging Face

Find the right model, dataset, or Space on the Hugging Face Hub and run Gradio apps as tools.

Book a demo View documentation

Tools

Tools discovered when you connect

Search the tool definitions captured from the MCP server. The live list can vary with version, deployment, and credentials.

Tools load when your organization connects this server

Hugging Face can expose tools based on its deployed version and credentials. Runlayer discovers the available list when your organization connects the server.

Compatibility

Connect Hugging Face to the clients your teams already use

27 AI clients can connect to Hugging Face through Runlayer. Pick one for its setup steps and controls.

Why Runlayer

Why connect Hugging Face through Runlayer?

One approved path

Approve Hugging Face once and everyone gets it in the AI clients they already use. Runlayer holds the credentials, so nothing sits in client configuration.

Fine-grained policy

Control who can use each server, tool, and resource, and under which runtime conditions.

Runtime security

Scan tool inputs and outputs for prompt injection, credential exposure, PII, and data exfiltration, then alert, mask, or block.

Audit and observability

Trace every request across users, clients, tools, policies, scanner decisions, and downstream systems.