AI and ML platforms - official Replicate server
Replicate MCP server
Exposes Replicate's HTTP API to agents so they can search, compare and fetch metadata for models, run predictions and retrieve results. Available as a hosted remote server or a local npm package. Replicate hosts it at https://mcp.replicate.com/sse, or you can run replicate-mcp (npm) locally with OAuth or API key auth.
Last verified 3 tools4 sources
Install
Claude Code
About Claude Codeclaude mcp add replicate https://mcp.replicate.com/sse --transport sse --scope userCodex
About Codexcodex mcp add replicate --env REPLICATE_API_TOKEN=r8_xxxxxx -- npx -y replicate-mcp@latestCursor
About Cursor{
"mcpServers": {
"replicate": {
"command": "npx",
"args": ["-y", "mcp-remote@latest", "https://mcp.replicate.com/sse"]
}
}
}VS Code
{
"servers": {
"replicate": {
"command": "npx",
"args": ["-y", "replicate-mcp"],
"env": {
"REPLICATE_API_TOKEN": "your-token-here"
}
}
}
}Generic
{
"mcpServers": {
"replicate": {
"command": "npx",
"args": ["-y", "replicate-mcp"],
"env": {
"REPLICATE_API_TOKEN": "your-token-here"
}
}
}
}Tools
search_models- Search for models (models.search in the HTTP API).
create_predictions- Run models by creating predictions (predictions.create in the HTTP API).
list_hardware- List available hardware options.
As documented on October 17, 2026. A running server reports its current list through tools/list.
Notes and caveats
- The remote server is recommended by Replicate; adding it opens a web-based auth flow where you paste a Replicate API token.
- Local server uses the REPLICATE_API_TOKEN env var and supports all HTTP API operations.
- Experimental Code mode (replicate-mcp@alpha --tools=code) needs Deno and runs locally only.
- License is from the npm package metadata; no public source repo is linked from the package.
- Only three tool names are shown in Replicate's docs (in the Codex example).
More MCP servers for AI and ML platforms
- AssemblyAI
AssemblyAI docs MCP server
Hosted MCP server that lets coding agents search and fetch AssemblyAI documentation pages on demand while they write integration code.
Transport: Streamable HTTPAuth: No auth - BasetenStatus: GA
Baseten MCP server
Remote backend MCP server at api.baseten.co/mcp that wraps the Baseten REST API (deployments, autoscaling, logs, secrets, training jobs and more) so coding agents can deploy, debug and operate models in a Baseten workspace. It acts as the signed-in user with that account's permissions.
Transport: Streamable HTTPAuth: OAuth or API key - Deepgram
Deepgram CLI MCP server
The dg CLI (deepctl) includes a built-in MCP server that proxies tools from the Deepgram developer API so AI coding tools can call Deepgram speech-to-text, text-to-speech and voice agent APIs. Runs over stdio by default or SSE.
Transport: stdio, SSEAuth: API key - ElevenLabs
ElevenLabs hosted MCP server
Remote MCP server that lets AI assistants create, configure and manage ElevenAgents voice agents in a workspace (conversations, tests, tools, knowledge base, branches, triage tickets) and generate speech audio. Authentication is OAuth with scoped workspace permissions.
Transport: Streamable HTTPAuth: OAuth - ElevenLabs
ElevenLabs MCP server
Local Python MCP server for ElevenLabs Text to Speech and audio APIs (speech generation, voice cloning, transcription, audio processing), run with uvx and an ElevenLabs API key.
Transport: stdioAuth: API key - Hugging Face
Hugging Face MCP server
Hosted MCP server that connects assistants to the Hugging Face Hub to search and explore models, datasets, Spaces, papers and documentation, run Jobs and sandboxes, and call MCP-compatible Gradio Spaces as tools.
Transport: Streamable HTTP, stdioAuth: OAuth or API key
Related reading
- What's New in Mastra (Late 2026): Harness, AgentController GA, Eval Gates and MCP 2.xMastra in late 2026: goals, schedules, signals, AgentController GA, eval gates and MCP 2.x, with type-checked TypeScript examples and honest tradeoffs.
- MCP vs A2A vs Function Calling: Which Layer Does What in 2026MCP vs A2A vs function calling compared on the 2026-07-28 MCP spec and A2A 1.0: what each protocol does, where they overlap, and how to stack them.
- Pydantic AI V2 in Late 2026: Capabilities, Graphs, Evals and MCPPydantic AI V2 explained through v2.57.0: capabilities, typed output, Conversation storage, caching, graphs, evals, MCP Apps and Logfire, with checked code.
- Reverse Engineer Anything (REA): What the MCP Server Does and Where the Line IsREA (Reverse Engineer Anything) is an MIT-licensed MCP server that lets Claude Code, Codex or Cursor inspect binaries, Electron apps and APKs. How to try it.
- Claude Opus 5.5 Built a Browser 3D World: The Higgsfield MCP WorkflowClaude Opus 5.5 and Higgsfield MCP built an interactive 3D Library of Alexandria in the browser. The verified connector setup and where the workflow holds up.
- Pi 1.0 Release Guide: MCP, Codemode and Pi DurablePi 1.0 adds built-in MCP through Codemode, a fullscreen TUI by default and Pi Durable. What changed, how to install it, and what breaks on upgrade.
FAQ
What does the Replicate MCP server do?
Exposes Replicate's HTTP API to agents so they can search, compare and fetch metadata for models, run predictions and retrieve results. Available as a hosted remote server or a local npm package. The vendor documents 3 tools, including search_models, create_predictions and list_hardware.
How do I add the Replicate MCP server to Claude Code?
Run claude mcp add replicate https://mcp.replicate.com/sse --transport sse --scope user, copied from Replicate's docs. This page also has snippets for Codex, Cursor, VS Code and Generic.
Does the Replicate MCP server need authentication?
Yes, OAuth or API key. Transport: SSE, stdio. Status: Unlabelled, last checked October 17, 2026.
Sources
Last verified against 4 Replicate pages, then re-checked independently.
- replicate.comhttps://replicate.com/docs/reference/mcp
- mcp.replicate.comhttps://mcp.replicate.com/
- registry.npmjs.orghttps://registry.npmjs.org/replicate-mcp
- github.comhttps://github.com/replicate/replicate-mcp-code-mode
Newsletter
New MCP servers, the week they ship
One email per week. New official MCP servers, the coding tools that support them and how developers wire them up.