# Replicate

> Replicate is a freemium MCP server for ai models and llm tooling. MCP server giving AI clients access to Replicate models, predictions and deployments. It can run locally on your machine or as a hosted remote server.

- **Kind:** MCP server
- **Pricing:** Freemium
- **Starting price:** $49 a month
- **Platforms:** API, CLI, self-hosted
- **Website:** https://pipeworx.io/packs/replicate
- **Repository:** https://github.com/pipeworx-io/mcp-replicate
- **MCP Registry name:** io.github.pipeworx-io/replicate
- **Categories:** [AI models and LLM tooling](https://ai-powered.com/categories/ai-and-llm)
- **Listing updated:** 2026-11-18

## About

The README describes this as an unofficial bridge to Replicate's HTTP API, run through the Pipeworx gateway. It exposes tools to browse public models, their versions and collections, search the catalog, and start model runs that are then polled for output or cancelled. Deployments, account details and the hardware list are available as well.

The server is a hosted endpoint at gateway.pipeworx.io, and it can also be run locally with npx as a stdio bridge that forwards calls to the same gateway. An optional Pipeworx API key raises the daily call cap; without one, calls are anonymous and limited. The README notes that the endpoint also lists shared gateway meta-tools such as ask_pipeworx alongside the pack's own tools.

It suits developers who want an agent to run open models such as Flux, Llama or Whisper on Replicate without writing their own API wrapper.

## Features

- **Model catalog browsing:** List public models, fetch model details, and page through version histories and schemas.
- **Keyword and collection search:** Search models by keyword or open curated collections such as text-to-image.
- **Prediction runs:** Start model runs from a version id and input, then poll status and output until complete.
- **Prediction management:** List recent predictions and cancel running ones.
- **Dedicated deployments:** List deployments and run predictions through them for faster cold starts.
- **Account and hardware lookup:** Fetch account details and the available CPU and GPU hardware types.

## Good for

- Generate images or audio from an agent by running a Replicate model and polling for output
- Find a suitable open model for a task by searching models or collections
- Check the input schema of a model version before constructing a prediction request
- Monitor or cancel long-running predictions from an AI assistant
- Pick a hardware SKU before creating a deployment

## Install

### Claude Code (Terminal)

```sh
claude mcp add --transport http replicate https://gateway.pipeworx.io/replicate/mcp
```

### Claude Desktop (Settings → Connectors → Add custom connector)

```sh
https://gateway.pipeworx.io/replicate/mcp
```

### Cursor (~/.cursor/mcp.json)

```json
{
  "mcpServers": {
    "replicate": {
      "url": "https://gateway.pipeworx.io/replicate/mcp"
    }
  }
}
```

### VS Code (.vscode/mcp.json)

```json
{
  "servers": {
    "replicate": {
      "type": "http",
      "url": "https://gateway.pipeworx.io/replicate/mcp"
    }
  }
}
```

### Windsurf (~/.codeium/windsurf/mcp_config.json)

```json
{
  "mcpServers": {
    "replicate": {
      "serverUrl": "https://gateway.pipeworx.io/replicate/mcp"
    }
  }
}
```

### Other clients (your client's MCP settings (most use this shape))

```json
{
  "mcpServers": {
    "replicate": {
      "type": "http",
      "url": "https://gateway.pipeworx.io/replicate/mcp"
    }
  }
}
```

## Tools

- `list_models`: List public models
- `model`: Model detail
- `model_versions`: List versions of a model
- `model_version`: Version detail with schema
- `search_models`: Search models by keyword
- `collections`: List model collections
- `collection`: Collection detail
- `create_prediction`: Start a model run
- `prediction`: Poll a prediction
- `list_predictions`: List recent predictions
- `cancel_prediction`: Cancel a prediction
- `list_deployments`: List your deployments
- `deployment`: Deployment detail
- `create_deployment_prediction`: Run via a deployment
- `account`: Your account details
- `hardware`: Available hardware types

## FAQ

### Is this an official Replicate integration?

No. The README says it is an independent, unofficial integration and is not affiliated with or endorsed by the upstream provider.

### Do I need an API key?

No. Calls work anonymously with a daily cap. The README says a free Pipeworx key raises the cap to 200 calls a day.

### How do I get the output of a run?

Start the run with create_prediction, which returns immediately, then poll the prediction tool until status is succeeded or failed, or pass a webhook URL.

### Is Replicate free?

Replicate has a free plan, with paid plans for more usage or features. Paid plans start at $49 a month.

### How do I add Replicate to Claude Code?

Run this command in your terminal: claude mcp add --transport http replicate https://gateway.pipeworx.io/replicate/mcp

### What are the best alternatives to Replicate?

The closest alternatives to Replicate are Epoch Ai Models, MCP, RelaySwitch, Fireworks AI MCP by usefulapi and Riftrunner AI MCP.

## Where this comes from

- Description: written from the project's README, the project's website and its MCP Registry entry.
- Install steps: from the official MCP Registry entry.

## Alternatives

- [Epoch Ai Models](https://ai-powered.com/mcp/epoch-ai-models): Epoch Ai Models is a freemium MCP server for ai models and llm tooling.
- [MCP](https://ai-powered.com/mcp/mcp-mwi57hue): MCP is a paid MCP server for design and media.
- [RelaySwitch](https://ai-powered.com/mcp/relayswitch-2): RelaySwitch is a paid MCP server for ai models and llm tooling.
- [Fireworks AI MCP by usefulapi](https://ai-powered.com/mcp/fireworks-ai-mcp-by-usefulapi): Fireworks AI MCP by usefulapi is a freemium MCP server for ai models and llm tooling.
- [Riftrunner AI MCP](https://ai-powered.com/mcp/riftrunner-ai-mcp): Riftrunner AI MCP is an open-source MCP server for ai models and llm tooling.
- [fal.ai MCP by usefulapi](https://ai-powered.com/mcp/fal-ai-mcp-by-usefulapi): fal.ai MCP by usefulapi is a freemium MCP server for ai models and llm tooling.
- [apimodels](https://ai-powered.com/mcp/apimodels): apimodels is a paid MCP server for design and media.
- [Synap](https://ai-powered.com/mcp/synap): Synap is a freemium MCP server for ai models and llm tooling.

---
Source: https://ai-powered.com/mcp/replicate
