
by Replicate
Replicate lets you run machine learning models through a cloud API without managing infrastructure. You can use public models, fine-tune with your own data, or deploy custom models.
APIMCP serverUsage-based
Add this server URL to an MCP-compatible client.
Client-specific setup is in Replicate’s documentation
Step 1: Sign up and create an API token.
Step 2: Pick a public model and read its page for inputs and a cost estimate.
Step 3: Call it with the Python or Node.js client, or the HTTP API.
Step 4: Add webhooks or streaming to receive results as they are produced.
Step 5: To use your own model, package it with Cog, push it, and optionally create a deployment to control hardware and scaling.
For public models you pay only for time spent processing your requests; setup and idle time are free. Most private models run on dedicated hardware, so you pay for setup, idle and active time.
Package it with Cog, Replicate's open-source tool, and push it to the platform. You can then run it as a private model or set up a deployment.
By default, public model requests go into a shared queue, so you may see cold boots or scaling limits. A deployment gives you your own instances and request queue.
Its docs list an MCP server and agent skills under the reference section.
Environment variables
Render NetworkRender Network is a decentralized GPU rendering platform that runs 3D and generative AI imaging jobs on GPU power contributed from around the world. Artists upload a scene, set the job options and download the finished frames.
SplineSpline is a browser-based 3D design tool where teams and AI agents model, animate and publish interactive 3D together in real time. Scenes can ship to the web, iOS and Android.