ChatGPT + Hugging Face: What the Integration Does in 2026
Yes, ChatGPT integrates with Hugging Face — there’s an official Hugging Face app in ChatGPT for inspecting models, datasets, Spaces, and papers. Behind it sits Hugging Face’s official hosted MCP server at huggingface.co/mcp, launched June 6, 2025 and open source — and HF’s docs explicitly list ChatGPT as a supported client. The server ships 7 built-in tools, almost all read-oriented, with one standout twist: you can attach any MCP-compatible Gradio Space as extra tools, which turns the Hub’s app ecosystem into ChatGPT abilities. As always, though, it runs in a session you’re driving — nothing watches your models or jobs between chats.
Here’s what the ChatGPT Hugging Face integration actually does, how to set it up, and what to use when you want ML-ops work that runs without you.
What ChatGPT can actually do with Hugging Face
The hosted MCP server exposes 7 built-in tools:
- Semantic search across the Hub. Find models, datasets, Spaces, and papers by describing what you need — “find me small instruction-tuned models suitable for on-device summarization.”
- Search the docs. Ask questions against Hugging Face’s documentation instead of tab-hopping.
- Inspect repo details. Pull the specifics of a model or dataset repo — what it is, how it’s licensed, what’s in it.
- Run & Manage Jobs. The one write-capable tool in the set — kick off and manage Jobs from the conversation.
- Attach Gradio Spaces as tools. The unique part: any MCP-compatible Space can be bolted on as an extra tool. Hugging Face’s own example — “Create a 1024x1024 image of a cat ghibli style” — runs an image-generation Space through the connection. Configure which Spaces per client at huggingface.co/settings/mcp.
How to set it up
- Have a ChatGPT plan where apps are available, plus a Hugging Face account.
- In ChatGPT, open Settings → Apps, find Hugging Face, and enable it — or add
https://huggingface.co/mcpas a custom connector (Business/Enterprise, or developer mode on Pro/Plus). - Authenticate with your HF token.
- Optionally, visit huggingface.co/settings/mcp to tune the per-client config and attach Gradio Spaces you want available as tools.
- Ask away — model search, dataset comparisons, paper lookups, or a Space invocation.
If your team lives in a coding agent instead, the same hosted server works there too — pointing Codex or another MCP-capable agent at huggingface.co/mcp gives it the identical toolset while you’re writing training code.
The limits that actually matter
- It doesn’t run on triggers. There’s no “when a Job finishes, summarize the results” or “when a new model drops in this niche, evaluate it against ours.” ChatGPT queries the Hub when you prompt it — it never fires on an event.
- Mostly read-only. Of the 7 built-in tools, only Run & Manage Jobs writes anything. Search, docs, and repo inspection are lookups.
- Session-bound, even in agent mode. ChatGPT Work runs are long and autonomous but manually started and metered — an errand, not a standing watch over your training runs.
- Cross-stack follow-through stops at the chat. The app won’t take the model comparison it produced and post it to Slack, log it in Notion, or email the team.
If you want ML team ops that run on their own: Carly
ML teams have a rhythm of recurring, event-shaped work: the weekly “what’s new on the Hub in our space” sweep, the job-completion report, the dataset-update notice that should reach the team without anyone asking. A chat session can’t own any of it.
That’s where Carly fits. Carly is an AI executive assistant that acts on triggers across your whole stack, and it natively integrates with Hugging Face:
- Fires on schedules and events, 24/7, in the cloud. “Every Friday, search the Hub for new open-weight models in speech recognition and email the team a shortlist with licenses” — done, weekly, unprompted.
- No-code setup by conversation. Describe the workflow in plain English; Carly interviews you and builds it.
- Connects Hub activity to the rest of your work — findings flowing into email, Slack, Notion, and task tools in one flow.
- Actually sends — drafts and sends email across Gmail and Outlook, so the model digest reaches inboxes, not a chat history.
- Connects to anything — 200+ native integrations, plus any other tool via your own API key.
AI agents start at $35/month, and steps in a workflow that don’t use AI run free and unlimited.
ChatGPT vs Carly
| ChatGPT (Hugging Face app) | Carly | |
|---|---|---|
| Semantic model/dataset/paper search | Yes | Yes |
| Run a Gradio Space from a prompt | Yes | Via your workflows |
| Weekly new-model digest, unprompted | No | Yes, on a schedule |
| Reacts to Hub or job events by itself | No | Yes, on any trigger |
| Runs without a session open | No (agent runs are started + metered) | Yes (cloud, 24/7) |
| Emails findings to the team | No | Yes (Gmail + Outlook) |
| Pushes results into Slack / Notion / tasks | No | Yes |
| Setup | Enable the app | Describe it in plain English |
| Pricing | Paid ChatGPT plan | AI agents from $35/mo |
ChatGPT’s Hugging Face app is a research surface you drive in a chat. Carly is an assistant that keeps your team current on the Hub while everyone’s heads-down training.
Frequently Asked Questions
Does ChatGPT work with Hugging Face?
Yes. Hugging Face has an official app in the ChatGPT directory, backed by its hosted MCP server at huggingface.co/mcp (launched June 6, 2025). Hugging Face’s docs explicitly list ChatGPT as a supported client. Enable it under Settings → Apps, authenticate with your HF token, and you can search models, datasets, Spaces, and papers conversationally.
What tools does the Hugging Face MCP server give ChatGPT?
Seven built-in tools: semantic search for models, datasets, Spaces, and papers; docs search; repo details; and Run & Manage Jobs (the only write-capable one). You can also attach any MCP-compatible Gradio Space as an extra tool via huggingface.co/settings/mcp — for example, generating images through an attached Space.
Can ChatGPT monitor Hugging Face for new models automatically?
No. The integration only runs inside a session you start — it can’t watch the Hub for new releases or fire when a Job finishes. For “when something changes, tell the team,” you need a trigger-based assistant like Carly, which connects natively to Hugging Face and runs on schedules and events.
How do I connect ChatGPT to Hugging Face with MCP?
Either enable the official Hugging Face app under Settings → Apps, or add https://huggingface.co/mcp as a custom connector on a plan that supports it. Both authenticate with your Hugging Face token, and per-client tool configuration lives at huggingface.co/settings/mcp.
More: Claude + Hugging Face · ChatGPT MCP · Codex + GitHub · ChatGPT agent mode · What is ChatGPT Work · Best AI agents for productivity
Ready to automate your busywork?
Carly schedules, researches, and briefs you—so you can focus on what matters.
See what people say
"Before Carly, I relied on a Calendly link, but the whole process felt impersonal and not very professional. Carly changed that by handling all the back-and-forth, so I'm no longer stuck in endless email threads trying to line up schedules.
Now Carly reaches out to candidates, shares my real-time availability, lets them pick a slot, then sends a Zoom link and drops it straight into my calendar. She sends reminders to both of us before each call, which has significantly reduced no-shows and last-minute confusion.
On top of scheduling, Carly acts like a full executive assistant, sending me my schedule the night before so I can prepare for each call. It reminds me of the old x.ai assistant, but Carly is noticeably smarter, faster, and better suited to my healthcare recruitment business."
