Find and Run AI Models on Replicate Through Your Muse

Through an AnythingMCP custom connector, your muse can find open AI models on Replicate, check what each one takes and costs, and run the one you pick to make images, video, audio, or text, billed to your Replicate account.

What it does

The connector adds 17 tools. Your muse can search public models by task or name, browse collections such as text-to-image or speech, and list the newest models. For one model, it can read the input fields, the latest version, and older versions. It can run an official model, run any model by its exact version, and run your own private deployments, follow the run, and list your recent runs. A run longer than about 25 seconds comes back in progress, and the muse checks it until it is done or you ask it to cancel.

How to use it

1. Sign in to Replicate, add a payment method or prepaid credit under Billing, then create an API token.
2. Start an AnythingMCP account, install the Replicate connector, and paste the token where it asks. It is sent only to Replicate.
3. In Muse, open Settings, then Connectors, and add a custom connector with the address AnythingMCP gives you. Sign in and approve.
4. Ask for the task, for example: find a model for upscaling old photos and tell me which inputs it takes.
5. Before any run, read the model name and rough cost, say yes, and save the result when it comes back.

What you need

• A Muse account that can add custom connectors
• A Replicate account with a payment method or prepaid credit, and an API token
• An AnythingMCP account with the Replicate connector installed

What to watch for

• Searching and reading models and runs are free. Running a model costs money. Some official image models cost a fraction of a cent per image, and video models can cost several dollars per run.
• The muse tells you which model will run and what it costs, and waits for your yes. It does not start run after run without asking. A run that fails is not billed.
• Download what you want to keep. Result links expire one hour after the run, and Replicate deletes inputs, outputs, and logs after one hour for API runs.
• The connector does not stream output, upload files, train models, or create or change models and deployments.
• This is a custom connector, and Meta does not review custom connectors. You can disconnect it in Settings at any time. The muse prepares, you decide.