# Replicate MCP for AI Agents AI Agent Connect > Replicate MCP lets your AI client run, search, and manage thousands of open-source machine learning models. Instead of worrying about local hardware or complex API setups, you can just tell your agent to generate an image, start a video, or find a specific model. It turns Replicate into a command line for your AI. ## Overview - **Category:** superpower - **Price:** Free - **Endpoint:** https://edge.vinkius.com/vk_preview_ZjGhumvOdiFG6C4qZnstV0k6xajsQmuKnK85lhvV/ai-agent-connect - **Tags:** machine-learning, model-inference, open-source-models, fine-tuning, api-access, generative-ai ## Description This Connector connects your AI to a massive library of open-source machine learning models. This integration lets you move from a text prompt to a finished video or a high-res image without ever opening a terminal or setting up a GPU environment. You can move your workflows into the cloud, letting your agent act as a production manager for machine learning tasks. You can ask it to find the best model for a specific job, check the requirements, and then kick off the job while you keep working on other things. If a process takes too long or you change your mind, you can just tell the agent to stop it. It makes it much easier to chain different types of AI outputs together, like generating an image and then passing that result into a video generation model. This is one of the standout tools in the Vinkius catalog for anyone trying to build actual products using open-source models. You get the power of a massive model library with the simplicity of just talking to your agent. Instead of wrestling with drivers or trying to find the right model version on a website, you can just describe the outcome you want. Your agent handles the heavy lifting of searching, verifying parameters, and managing the execution status. It turns a complex technical hurdle into a simple conversation. You no longer need to worry about hardware limitations or the nuances of different model versions. Whether you are looking for a specific audio codec or a high-end image generator, the agent finds it for you and starts the work immediately. This gives you a way to iterate on creative projects at a much faster pace. ## Tools ### create_prediction Start a new model task by sending a JSON payload of inputs to the AI. This is how you actually trigger a generation. ### get_account Pull your current account details to make sure everything is connected correctly. Use this to verify your credentials. ### get_collection Pull the full list of models for a specific category like 'text-to-image'. Use this to see all options in a niche. ### get_model Get the exact schema and requirements for a specific model to ensure your inputs are correct. This prevents errors before you start. ### get_prediction Check the status and see the final output of a specific running task. Use this to monitor progress as it happens. ### list_collections Browse curated groups of models like 'Audio Generation' or 'Image-to-Text'. This helps you find models by category. ### list_predictions View a history of your recent model requests and their results. This is great for auditing your previous work. ### search_models Find specific models using keywords like 'TTS' or 'Stable Diffusion'. This is the fastest way to discover new tools. ### list_models See every public model available on the Replicate platform. This helps you understand the scope of what you can run. ### list_deployments View your active model deployments to see what's currently running. This is useful for tracking ongoing infrastructure. ### cancel_prediction Stop a running model task immediately to prevent unnecessary costs. This gives you instant control over your spending. ### list_hardware See what GPU options are available for different types of model runs. This helps you understand the underlying compute. ## Prompt Examples **Prompt:** ``` I need to make a high-quality video of a dragon flying over a mountain. Can you find a model for that and start it? ``` **Response:** ``` I've searched the Replicate library for high-quality video generation models. I found a highly-rated model that handles cinematic prompts well. **Model Selected:** `dream_video_generator` **Parameters:** - **Prompt:** a dragon flying over a mountain - **Resolution:** 720p - **Duration:** 5s I am starting the prediction now. You can ask me for the status at any time. ``` **Prompt:** ``` What are some of the best open-source models for text-to-speech right now? ``` **Response:** ``` Here are some of the top-rated open-source models for TTS currently available on Replicate: | Model Name | Best Use Case | | :--- | :--- | | **Bark** | Realistic speech with non-verbal sounds | | **XTTS v2** | High-quality cloning and multi-language support | | **Coqui TTS** | Versatile, high-speed generation | Would you like me to check the specific requirements for any of these? ``` **Prompt:** ``` Show me the status of my last few image generations. ``` **Response:** ``` I've pulled your recent prediction history. Here is the status of your last 3 requests: * **Request `p_9821...`**: ✅ Finished (Image of a cybernetic forest) * **Request `p_9825...`**: ⏳ In Progress (80% complete - Abstract oil painting) * **Request `p_9830...`**: ❌ Cancelled (User terminated early) ``` ## Capabilities ### Run ML predictions Start image, video, or audio generation tasks via natural language. ### Search for models Find specific open-source models based on keywords like 'TTS' or 'Stable Diffusion'. ### Monitor active jobs Check the status of long-running model tasks in real time. ### Cancel running tasks Stop a prediction instantly to prevent unnecessary costs. ### Explore model collections Browse curated groups of models like text-to-image or audio generation. ### View execution history See a log of your recent model requests and their results. ## Use Cases ### Video Generation Ask your agent to make a 5-second clip of a dragon flying over a mountain. It finds a video model, checks the schema, and starts the run. ### Model Comparison Ask your agent to find the best open-source models for high-quality speech. It searches, lists collections, and compares options for you. ### Batch Image Work Tell your agent to generate 10 different styles of cats. It creates multiple predictions and monitors them all at once. ### Workflow Integration Tell your agent to create an image, then use a different model to turn it into a 3D object. It chains multiple create_prediction calls. ## Benefits - Skip the GPU setup by running everything in the cloud via create_prediction. - Save money on long tasks by using cancel_prediction the moment you see an error. - Find the right tool for the job faster by using search_models and list_collections. - Ensure your inputs work the first time by checking schemas with get_model. - Track your history and costs easily by viewing logs with list_predictions. ## How It Works The bottom line is you get a direct pipeline from natural language to production-grade machine learning outputs. 1. Get your Replicate API token from your account settings and add it to your configuration. 2. Connect the Connector to your preferred AI client like Claude or Cursor. 3. Tell your agent what you want to generate or find, and it handles the rest. ## Frequently Asked Questions **How does Replicate MCP help with video generation?** It lets your AI agent find the best video generation models and start the process for you. You just describe the video you want, and the Connector handles the technical work of running it in the cloud. **Can I use Replicate MCP to run open-source models?** Yes, this Connector gives your AI client access to thousands of open-source models. You can run anything from image generators to specialized language models without needing your own hardware. **Does Replicate MCP save me from managing my own hardware?** Exactly. It offloads all the heavy lifting to the Replicate cloud. You don't need a GPU or a complex setup; your agent handles the inference for you. **How do I find specific models using Replicate MCP?** You can just ask your AI agent to search for a specific capability, like 'text to speech' or 'face swap'. The Connector will scan the platform and show you the best matches. **Can my AI agent stop a running task with Replicate MCP?** Yes, your agent can cancel any running prediction instantly. This is helpful if you notice an error early on or just want to stop a task to save on costs. **How does Replicate MCP handle model parameters?** The Connector can check the exact requirements for any model. This ensures that when your AI agent starts a task, it uses the correct settings and variables every time. **Can the agent pass a JSON payload directly into a Replicate model?** Yes. You can utilize the `create_prediction` action and attach the payload parameter filled out with any required input schema (e.g., specific `prompt`, `num_inference_steps`). Since models change inputs constantly, you should always ask your assistant to fetch the schema details first via `get_model` to verify keys. **Does the prediction command return results instantly?** No, Replicate's API operates asynchronously. The initial command gives your assistant an ID. You must then ask your AI companion to query the `get_prediction` tool periodically using that generated ID until it displays the completed status along with the generated web URLs or generated strings. **Can the AI browse trending or curated model collections?** Yes. Use the `list_collections` tool to browse curated groups of models organized by category — such as image generation, text-to-speech, or video. Each collection includes a slug and description so you can quickly identify the right set of models for your use case.