replicate-core-openapi
Core Replicate REST API for discovering public models, inspecting model versions, creating and monitoring AI predictions, listing predictions, and canceling a running prediction. The connected user supplies a Replicate API token at connection time; no credential is included. Official documentation: https://replicate.com/docs/reference/http and https://replicate.com/docs/reference/openapi
Connect this API to an AI agent Browse all specs
What you can do with replicate-core-openapi via MCP
Every endpoint below becomes a governed tool your AI agent (Claude, ChatGPT, or a customer-facing VX agent) can call — with scoped permissions, sandbox testing, and audit logs.
| Method | Path | Operation | Description |
|---|---|---|---|
| GET | /account |
getAccount | Get the authenticated account |
| GET | /hardware |
listHardware | List available hardware |
| GET | /models |
listModels | List public models |
| GET | /models/{owner}/{name} |
getModel | Get a model and its latest version |
| GET | /models/{owner}/{name}/versions |
listModelVersions | List versions for a model |
| GET | /predictions |
listPredictions | List predictions |
| POST | /predictions |
createPrediction | Create a prediction |
Authentication
bearerAuth
Bearer token — sent as Authorization: Bearer <token>
Usage guide
Getting credentials
- Create or log in to a Replicate account at replicate.com.
- Open the API tokens page and create a personal API token.
- Keep the token private. Do not paste it into the Swagger file, public marketplace description, prompts, or chat messages.
- When VX Agent asks for authentication, provide the token using the Bearer authentication method.
How to use this API
Use the Replicate API to discover public AI models and run model predictions for image generation, image upscaling, speech-to-text, text-to-speech, video generation, OCR, background removal, and other supported AI tasks.
A typical workflow is:
- Use
GET /modelsorGET /models/{owner}/{name}to find and inspect a model. - Use
GET /models/{owner}/{name}/versionsto identify a model version and review its input schema. - Use
POST /predictionswith the selected version and the model-specificinputobject. - Use
GET /predictions/{prediction_id}to check whether the prediction isstarting,processing,succeeded,failed, orcanceled. - When the prediction succeeds, use the returned output value or URL. Save important output files promptly because provider retention rules may remove them later.
- Use
POST /predictions/{prediction_id}/cancelonly when the user explicitly asks to stop a running prediction.
The input fields are different for every model. Always inspect the selected model's input schema before creating a prediction. Images, videos, and audio files should normally be provided as HTTPS URLs or supported data URLs. Prediction requests may consume Replicate account credits or incur charges depending on the model and hardware.
Safety and usage notes
Require confirmation before starting an expensive, long-running, or potentially sensitive generation. Do not send private files to a model unless the user has authorized it. Do not expose API tokens or model outputs containing personal or confidential information. Do not delete models, deployments, versions, or other account resources through this connection unless the user explicitly requests it.
Official documentation: Replicate HTTP API and Replicate OpenAPI reference.
How to use this spec
- Create a free VX Agents account (sandbox tools included).
- Open the marketplace and connect “replicate-core-openapi” to an agent.
- Expose it as an MCP server for Claude/ChatGPT, or chat with it directly.