Replicate Alternative MCP Server
Run ML models via Replicate — generate images, text, audio and video from community models, track predictions and explore collections from any AI agent.
Ask AI about this MCP Server
Vinkius AI Gateway supports streamable HTTP and SSE.
Works with every AI agent you already use
…and any MCP-compatible client
What is the Replicate Alternative MCP Server?
The Replicate MCP Server gives AI agents like Claude, ChatGPT, and Cursor direct access to Replicate. Run ML models via Replicate — generate images, text, audio and video from community models, track predictions and explore collections from any AI agent. Powered by the Vinkius AI Gateway — no API keys, no infrastructure, connect in under 2 minutes.
Replicate MCP Server: see your AI Agent in action
Built-in capabilities (12)
cancel_prediction
Provide the prediction ID. The prediction status will change to "canceled". Cancel a running prediction
create_prediction
Requires the model slug in "owner/name" format and an input object matching the model's schema. Optionally specify a version ID and webhook URL. Returns the prediction object with its ID, status (starting, processing, succeeded, failed, canceled) and output. Use get_prediction to check status and retrieve results. Run a model prediction on Replicate
get_account
Returns account type, username and usage info. Use this to verify your API token is working correctly. Get the authenticated Replicate account info
get_collection
Provide the collection slug (e.g. "text-to-image", "large-language-models"). Get details for a specific model collection
get_model
Provide the model slug in "owner/name" format (e.g. "stability-ai/sdxl" or "meta/meta-llama-3-70b-instruct"). Get details for a specific Replicate model
get_model_versions
Each version includes its ID (64-char hash), creation date, input/output schema and cog version. Use this to find the correct version ID when creating predictions for models that require a specific version. Get all versions of a Replicate model
get_prediction
Returns the prediction ID, status (starting, processing, succeeded, failed, canceled), input, output URLs, creation time and logs. Use the prediction ID returned from create_prediction. Get the status and result of a prediction
list_collections
Collections group related models by category (e.g. "text-to-image", "large-language-models", "audio-to-audio", "image-to-video"). Each collection includes its slug, name, description and featured models. List model collections on Replicate
list_hardware
Each hardware option includes its SKU name, pricing and specifications. Useful for choosing the right GPU for your prediction workload. List available GPU hardware on Replicate
list_models
Each model includes its name, owner, description, run count, hardware requirements and cover image URL. Use this to discover available models for running predictions. List available ML models on Replicate
list_predictions
Each prediction includes its ID, model, status, creation time and output URLs. Useful for tracking prediction history and monitoring model usage. List recent predictions on Replicate
search_models
Returns models with their name, owner, description, run count and hardware. Useful for finding specific types of models (e.g. "text-to-image", "llm", "music-generation"). Search for models on Replicate by query
What this connector unlocks
Connect your Replicate account to any AI agent and run thousands of open-source ML models through natural conversation.
What you can do
- Model Discovery — Browse, search and inspect thousands of ML models with their descriptions, run counts and hardware requirements
- Predictions — Run models by creating predictions and tracking their status (starting, processing, succeeded, failed)
- Collections — Explore curated collections of models by category (text-to-image, LLMs, audio, video)
- Hardware Options — View available GPU types and pricing for model inference
- Account Info — Check your account details and usage
How it works
1. Subscribe to this server
2. Enter your Replicate API Token
3. Start running ML models from Claude, Cursor, or any MCP-compatible client
No more navigating the Replicate website to find models or check prediction status. Your AI acts as a dedicated ML operations assistant.
Who is this for?
- Developers — quickly run image generation, text models and other ML predictions via conversation
- ML Engineers — discover models, compare hardware requirements and track prediction history
- Researchers — explore model collections, inspect version schemas and test models before deployment
Frequently asked questions
Give your AI agents the power of Replicate
Access Replicate and 2,500+ MCP servers — ready for your agents to use, right now. No glue code. No custom integrations. Just plug Vinkius AI Gateway and let your agents work.