Replicate
Run Replicate models, deployments, and predictions from your agent while building AI features in your app.
Overview
com.mcparmory/replicate is a MCP server for the Build phase that runs Replicate models, predictions, and deployments through your coding agent.
What is this MCP server?
- MCP tools to run AI models and manage predictions via Replicate’s cloud API
- Supports creating and managing Replicate deployments from the agent session
- Stdio MCP via uvx mcparmory-replicate 1.0.2 or Docker ghcr.io/mcparmory/replicate:1.0.2
- Registry entry from MCP Armory (github.com/mcparmory/registry) with schema 2025-12-11
- Fits solo builders prototyping image, audio, or LLM pipelines without custom CLI scripts each time
- Server version 1.0.2
- 2 packages: PyPI identifier mcparmory-replicate and OCI Docker
- stdio transport per MCP registry manifest
Community signal: 25 GitHub stars.
What problem does it solve?
Builders copy API snippets and switch to Replicate’s UI to test models while the agent cannot start predictions or check deployment state.
Who is it for?
Indie developers integrating Replicate inference into apps or agents who want repeatable, chat-driven prediction and deployment ops.
Skip if: Teams training custom weights locally only, or products with zero cloud inference needs.
What do I get? / Deliverables
After registration, the agent invokes Replicate via MCP so model runs and deployment tasks stay in your development workflow.
- Agent-accessible Replicate model run and prediction tools
- Deployment management actions via MCP where exposed by the server
- Faster iterate loop on model choice during feature build
Recommended MCP Servers
Journey fit
Model runs and hosted deployments are core product integration work during Build, not distribution or ops monitoring. Replicate is an external API you wire into backends and agents—canonical shelf is build/integrations.
How it compares
MCP bridge to Replicate’s HTTP API, not an on-device inference skill or a generic Hugging Face downloader.
Common Questions / FAQ
Who is com.mcparmory/replicate for?
Solo builders and small teams using Replicate in production or prototypes who want Claude Code, Cursor, or Codex to run models and manage predictions via MCP.
When should I use com.mcparmory/replicate?
During Build when you are integrating inference, comparing model outputs, or tuning deployments and want the agent to execute Replicate API actions directly.
How do I add com.mcparmory/replicate to my agent?
Configure MCP stdio with uvx for mcparmory-replicate 1.0.2 or the published Docker image, and set your Replicate API token in the environment or secrets your client supports.