- EXECUTION
- MONITORING
- DELIVERY
- HUMAN
Short answer
How can I run ML models through my AI?
Your AI uses the Replicate MCP to send model inputs and monitor the status of the run. You'll get direct access to the output URLs once the model finishes. This breakdown shows how your agent handles the entire lifecycle from trigger to result.
The outcomes.
Where your model runs land.
The AI manages the heavy lifting of the execution and keeps you updated on the results.
EXECUTION
Model runs start
Your AI sends the parameters to Replicate. It handles the initial request and starts the process immediately.
MONITORING
Status tracking
The AI checks if the model is still processing or if it's finished. You don't have to manually refresh a dashboard.
DELIVERY
Output URLs ready
Once the run completes, the AI pulls the specific URL for your generated image or data. You get the link right in your chat.
HUMAN
Manual intervention
If a run hangs or you change your mind, you tell your AI to stop it. The AI handles the cancellation for you.
The workflow
What your AI does when a prediction is requested.
The AI acts as the bridge between your text instructions and the Replicate infrastructure.
Trigger the run
Your AI takes your description or parameters and sends them to the model to begin processing.
create_predictionCheck progress
The AI periodically asks for the current status to see if the model is still working.
get_predictionReview history
If you need to find a previous run or check what's currently active, the AI looks through your recent activity.
list_predictionsStop a run
If a model is taking too long or you made a mistake, the AI kills the process.
cancel_prediction
Try it
Copy these to start.
Paste these into your chat to begin working with Replicate.
Starting points
Replace the bracketed text with your specific model names or input data.
Replicate Connector
You're all set. Choose your MCP client and follow the setup instructions.
https://edge.vinkius.com/vk_preview_M4yBN2KNxSS8lU5z0YgsZ7knjRqsTTpfrLzeO7vW/mcpClaude Desktop
Follow the steps below to connect in seconds.
- 1In Claude Desktop, open Settings → Connectors.
- 2Click “Add custom connector” and paste the connector link above as the remote MCP server URL.
- 3Click Add and start a new chat — Replicate capabilities are ready to use.
{
"mcpServers": {
"replicate-alternative-mcp": {
"url": "https://edge.vinkius.com/vk_preview_M4yBN2KNxSS8lU5z0YgsZ7knjRqsTTpfrLzeO7vW/mcp"
}
}
}Run a prediction using the stable-diffusion model with the prompt [insert prompt here].
Check the status of my last prediction to see if the output URL is ready.
List my recent predictions so I can find the URL for the image I just made.
Cancel the prediction I just started for the [model name] model.
Claude
ChatGPT
Cursor
VS Code
Windsurf
Claude Code
JetBrains
Cline
Start here
Connect Replicate once, then ask.
Just click the link to connect your account. Your credentials stay encrypted and you'll be ready to run models immediately.
Connect Replicate to your AIFAQ
How this task behaves.
- 01
Can my AI change the model settings on Replicate?
No. The AI can only pass the parameters you provide in your prompt to the model.
- 02
Will the AI delete my previous prediction history?
No. The AI can list and view predictions, but it does not have a tool to delete them.
- 03
How do I get the image or file produced by the model?
The AI retrieves the output URL from the prediction once it reaches a finished state.
- 04
Can the AI start a run on its own?
No. The AI only triggers a prediction when you explicitly ask it to via your chat.
- 05
Can I stop a model that is already running?
Yes. You can tell your AI to cancel a specific prediction if it is still in progress.
More questions about Replicate? See everything the Replicate Connector can do
Connect Replicate to Claude, Cursor, ChatGPT & more
