search_available_modelsLABSCO SUMMARY
Generations you can come back to, not just receive.
Treating each generation as a resource is what separates this from a single image call — you can ask for the landscape ones you made last week without keeping the URLs yourself. The text-to-image path is the one the author calls thoroughly tested; treat the rest of the prompt set as less worn in.
What it is
A FastMCP server for the Replicate API focused on image generation. Rather than one fire-and-forget call, generations are resources you can list, search and check the status of, driven by prompts that handle style and quality for you.
What you get
- A text-to-image prompt with detailed style control, quality presets (draft, balanced, quality, extreme), size and aspect-ratio customisation, and progress tracking
- Resource-based generation management — browse recent generations, search past ones, and check the status of the last run
- Model discovery and selection help, plus a parameter-help prompt for configuring a model you have not used before
- Webhook integration for external notifications, and real-time updates through subscriptions
- Image-to-image transformation is listed as coming soon
Requirements
Python 3.11 or higher and a Replicate API key from your Replicate account. Install with uv pip install mcp-server-replicate (or uvx or pip), set REPLICATE_API_TOKEN in the environment or the client's env block, and register uv tool run mcp-server-replicate in claude_desktop_config.json. Package version 0.1.9. MIT.
Setup effort
One command plus a key — uv pip install mcp-server-replicate, then supply credentials
USE CASES
01
Making an image without learning a model's parameters
Describe what you want with a quality level and style, and let the prompt map it onto the model's settings.
02
Finding something you generated earlier
Search your past generations by description instead of hunting through saved files.
03
Picking the right model for a job
Ask for help choosing among Replicate's models, then get the parameters explained before you run it.
3 tools
Listed in project docs. Your client may show a different list.
get_model_detailsgenerate_imageBEFORE YOU INSTALL
What to know first
Tools you get
3 tools named in the project's documentation.
Requires
- uv on PATH
- Python 3.11
- REPLICATE_API_TOKEN in your environment
Where it runs
The documented client runs the server on your machine.
INSTALLATION
Connect your client
Server setup
TERMINAL
uv pip install mcp-server-replicateMCP Clients
Claude Desktop
CLAUDE DESKTOP · JSON
{
"mcpServers": {
"replicate": {
"command": "uv",
"args": [
"tool",
"run",
"mcp-server-replicate"
],
"env": {
"REPLICATE_API_TOKEN": "<YOUR_REPLICATE_API_TOKEN>"
}
}
}
}<YOUR_REPLICATE_API_TOKEN> is your Replicate API token. The README's sample also sets "cwd": "$PATH_TO_REPO" and a "globalShortcut"; neither is needed, because uv tool run fetches the published package, and the shortcut line would change Claude Desktop's own keyboard shortcut.
