Run OneInfer from Claude, ChatGPT, or Cursor

Connect your assistant to OneInfer over MCP, the open standard AI assistants use to call outside tools. Ask in plain language and it uses 24 OneInfer tools for models, images, video, Studio, GPU machines, and dedicated endpoints, all on your own account.

Set up your assistant Add to Claude

New to OneInfer? Create an account first.

Pick a request

You: Start a GPU machine with one H100 for my fine-tuning run. My Developer ID is 3f9c2a7e1b8d4c6fa0e57d21b94c8e13.

Tool call: oneinfer_list_gpusChecked which GPU types are available

Tool call: oneinfer_get_gpu_pricingLooked up the H100 price

Assistant: An H100 is available at $2.49 an hour, paid from your credits. Should I start it?

You: Yes, start it.

Tool call: oneinfer_create_gpu_machineStarted a machine with 1 H100

Assistant: Your H100 machine is starting. It bills by the hour until you stop it in Console.

Result: GPU machine

GPU: 1 × H100
Price: $2.49 an hour
Status: Starting

You: Make a product shot of a matte black speaker, then turn it into a 5-second clip.

Tool call: oneinfer_generate_imageGenerated a 1024 × 1024 image

Tool call: oneinfer_generate_videoStarted a 5-second 720p video from the image

Tool call: oneinfer_get_video_taskChecked the video job until it finished

Assistant: Both are ready. The clip starts from the product shot.

Result: speaker.png and speaker.mp4

Image: 1024 × 1024
Video: 5 seconds, 720p

You: Put Qwen3 30B on a dedicated endpoint with 2 H100s. My Developer ID is 3f9c2a7e1b8d4c6fa0e57d21b94c8e13.

Tool call: oneinfer_get_gpu_pricingLooked up the H100 price

Assistant: Two H100s come to $4.98 an hour, paid from your credits. Should I deploy it?

You: Yes, deploy it.

Tool call: studio_create_endpointCreated an endpoint on 2 H100s

Tool call: studio_get_endpoint_statusChecked the deployment

Assistant: Your endpoint is deploying. Ask me to check on it any time, and delete it in Console when you no longer need it.

Result: Dedicated endpoint

Model: Qwen3 30B
GPUs: 2 × H100
Status: Deploying

You: Make a 30-second launch film for our new voice API, with music and a voiceover.

Tool call: studio_quote_movieGot a credit quote for a 30-second movie

Result: Credit quote

Film: 30 seconds
Includes: Video, music, voiceover
Quote: About 370 credits
Charged: Only if you go ahead

Assistant: Here is the quote. Getting it was free. Should I make the film?

You: Yes, go ahead.

Tool call: studio_create_movieStarted the movie from your brief

Tool call: studio_generate_musicMade a 30-second music track

Tool call: studio_generate_voiceoverRecorded the voiceover script

Tool call: studio_stitch_movieCombined video, music, and voice

Result: launch-film.mp4

Length: 30 seconds
Tracks: Video, music, voiceover

Example sessions. The tool names are the ones your assistant shows when it calls OneInfer; prices, IDs, and results are examples.

What your assistant can do

The 24 tools fall into six areas. Your assistant picks the right ones from what you ask, so you never need the names.

See prices Find your Developer ID

Your Developer ID (optional)

It’s in Console under Settings and isn’t a secret. Paste it here to fill it into the prompts below.

Models and chat

The models on your OneInfer account.

Chat spends credits

See every model available on your account
Chat with any of them using text messages
Compare model prices

3 tools

oneinfer_list_models
oneinfer_get_model_pricing
oneinfer_chat_completion

Try asking

List the OneInfer models I can use, with their prices.

Images and video

Pictures and short clips from OneInfer’s image and video models.

Spends credits

Generate images from a prompt, at a chosen size or aspect ratio, or from a reference image
Turn a prompt or an image into a 2 to 30 second video at 480p, 720p, or 1080p
Check on videos while they render

3 tools

oneinfer_generate_image
oneinfer_generate_video
oneinfer_get_video_task

Try asking

Generate a 16:9 image of a lighthouse at dusk with a OneInfer image model.

Studio

OneInfer Studio turns a written brief into a finished film, with music and a voiceover.

Spends credits; quotes are free

Get a credit quote before you make a movie
Turn a written brief into a movie up to 10 minutes long and follow its scenes as they render
Make music and voiceovers, then stitch video, voice, and music into a final cut
Browse your Studio storage

10 tools

studio_quote_movie
studio_create_movie
studio_get_movie
studio_list_movies
studio_generate_music
studio_generate_voiceover
studio_get_audio_job
studio_stitch_movie
studio_list_storage
studio_get_storage

Try asking

Quote a 60-second OneInfer Studio movie about a city waking up.

GPU machines

Cloud machines with GPUs, for training runs and other long jobs. Stop them in Console when you’re done.

Starting one spends credits
Needs Developer ID

See which GPU types are available and what they cost
List your GPU machines
Start a new machine with up to 8 GPUs

4 tools

oneinfer_list_gpus
oneinfer_get_gpu_pricing
oneinfer_list_gpu_machines
oneinfer_create_gpu_machine

Try asking

List my OneInfer GPU machines. My Developer ID is your Developer ID.

Dedicated endpoints

A model served from GPUs that only you use, at its own address. Delete them in Console when you’re done.

Creating one spends credits
Needs Developer ID

Deploy a model to its own endpoint on 1 to 8 GPUs
List your endpoints and check their status

3 tools

studio_create_endpoint
studio_list_endpoints
studio_get_endpoint_status

Try asking

Show the status of my OneInfer dedicated endpoints. My Developer ID is your Developer ID.

Credits

Your OneInfer balance.

Free to check
Needs Developer ID

Check your credit balance

1 tool

oneinfer_get_credits

Try asking

How many OneInfer credits do I have? My Developer ID is your Developer ID.

Limits Video length, GPU counts, and request rates

Video: 2 to 30 seconds, 480p to 1080p
Movies: Up to 10 minutes
Music: 5 seconds to 10 minutes
Prompts and scripts: Up to 8,000 characters
GPU machines: Up to 8 GPUs
Dedicated endpoints: 1 to 8 GPUs
Tool calls: Up to 60 a minute
Each tool call: Up to 60 seconds

Videos and movies render in the background. Ask your assistant to check on them.

Sign-in and safety

The connector acts only as you, only after you sign in, and only through the 24 tools listed above.

How sign-in works

1
Your assistant connects to mcp.oneinfer.ai. There’s nothing to set up on OneInfer first.
2
You sign in on oneinfer.ai with your usual account.
3
Your assistant gets a sign-in token for your account. It lasts an hour and renews on its own. Your assistant never sees your password or an API key.

To disconnect, remove the OneInfer connector in your assistant’s settings.

What it can’t do

There are no tools to buy credits, create API keys, delete your account, or change admin settings.
Requests can’t carry an API key. Signing in is the only way in.
Reference images must be public web links. Links to private or internal networks are blocked.
Every request runs on your own account and nobody else’s.

It can start GPU machines and endpoints, but it can’t stop them. Stop machines from the Console home page and delete endpoints under Dedicated endpoints.

Set it up in your assistant

Pick your assistant and follow the steps. New to OneInfer? Create an account first.

Add OneInfer as a custom connector in one click.

1
Click Add to Claude. The connector name and URL are filled in for you.
2
Sign in with your OneInfer account when Claude asks.
3
Start a new chat and ask for anything on OneInfer.

Add to Claude

Then try this first

List the OneInfer models I can use, with their prices.

Stuck? Read the full setup guide.

You always sign in with your OneInfer account. Never paste an API key into the chat.

Using another MCP client?

Add this URL as a Streamable HTTP server and sign in with your OneInfer account when asked.

https://mcp.oneinfer.ai/mcp

Other ways to install

OneInfer is listed on Smithery and in the official MCP Registry.

Open on Smithery View in MCP Registry

Questions

What is MCP?

MCP (Model Context Protocol) is an open standard that lets AI assistants like Claude, ChatGPT, and Cursor use outside tools. The OneInfer MCP server is what connects your assistant to your OneInfer account.

Which assistants can I use?

Claude, Claude Code, ChatGPT, Perplexity, Codex, Cursor, and VS Code with GitHub Copilot have setup steps on this page. Any other MCP client that supports Streamable HTTP and sign-in works too, and you can install OneInfer from Smithery or find it in the MCP Registry.

Do I need to know the tool names?

No. Ask in plain language and your assistant picks the right tools. Most assistants show which tool they called, so you can see what happened.

Does using it spend credits?

Anything that makes something spends credits: chat, images, video, Studio movies, music, voiceovers, final cuts, new GPU machines, and dedicated endpoints. Looking things up, like model lists, prices, and your balance, doesn’t. Ask for a movie quote before you commit.

How do I stop a GPU machine or a dedicated endpoint?

Your assistant can start them but can’t stop them. Stop GPU machines from the Console home page, and delete dedicated endpoints under Console, then Dedicated endpoints.

What is my Developer ID, and is it safe to share?

It’s a 32-character code that identifies your OneInfer account, like 3f9c2a7e1b8d4c6fa0e57d21b94c8e13. It isn’t a secret, so you can paste it into your assistant. Find it in Console under Settings. Requests about credits, GPU prices, GPU machines, and dedicated endpoints need it.

How do I disconnect OneInfer?

Remove the OneInfer connector in your assistant’s settings.

Start with one request

Add OneInfer to your assistant, sign in, and ask it to list the models you can use.

Add to Claude Set up another assistant