# genrelay-mcp [Health: Active]

**Category:** 🎥 Multimedia Process  
**Repository:** https://github.com/gaoruitech/genrelay-mcp  
**GitHub Stars:** 0  
**npm Downloads (last month):** 470  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/genrelay-mcp

## Description
Generate video and images (Veo 3.1, Grok Imagine, Nano Banana, GPT Image 2) via the GenRelay API.

## Tools
Capabilities this server exposes over MCP:

- **generate_video** — Text-to-video and image-to-video. Returns the clip, or a task id if it's still running
- **generate_image** — Image generation and editing. Returns the picture **inline
- **check_job** — Pick up a job by task id
- **list_models** — What's available and what each model is for

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "genrelay-mcp": {
    "command": "npx",
    "args": ["-y","genrelay-mcp"],
    "env": {
      "GENRELAY_API_KEY": "",
      "GENRELAY_BASE_URL": ""
    }
  }
}
```

**Requires environment variables:** `GENRELAY_API_KEY`, `GENRELAY_BASE_URL` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation

## What genrelay-mcp MCP server does

genrelay-mcp MCP server exposes GenRelay’s media-generation API through the Model Context Protocol. It is intended for MCP clients that need to create video or image assets from natural-language requests, rather than for managing a local media-processing pipeline.

The available model families include Google Veo 3.1, xAI Grok Imagine, Omni Flash, Nano Banana Pro and 2, and GPT Image 2. Video options cover text-to-video, image-to-video, first/last-frame generation, component-guided generation, and editing from a source video where the selected model supports that operation. Image models support generation and editing.

This package fits the Multimedia Process category because its tools are focused on creating and handling generated visual media. Model selection is exposed through the server, so an agent can inspect available models and their intended uses before submitting a request.

## How it works

The server runs as a standard stdio MCP process launched with `npx`. An MCP client sends tool calls to the process, which forwards generation requests to GenRelay using the configured API key.

Image requests generally complete within the tool-call time budget and return the generated picture inline. Video generation can take one to five minutes, so `generate_video` may return a task ID and an in-progress status instead of the finished clip. The generation continues on the server side. Calling `check_job` with that task ID retrieves the job result; submitting the same video request again can create another billable job.

Model output tiers include 1k, 2k, and 4k for images, plus 720p, 1080p, and 4k for Veo video. Pricing is tier-based. If the requested size does not match a supported tier, the model’s base rate is used, so callers should set the tier explicitly when cost selection matters.

## Setup and configuration

No package installation is required for the client setup. Add an MCP server entry that runs `npx -y genrelay-mcp`, then provide a GenRelay API key through the environment. A key can be created through the GenRelay console.

`GENRELAY_API_KEY` is required. `GENRELAY_BASE_URL` is optional and defaults to `https://genrelay.ai/v1`; it can be used when a different GenRelay API endpoint is needed. Restart the MCP client after changing its configuration.

Claude Desktop uses `claude_desktop_config.json`, while Cursor uses `.cursor/mcp.json`. Both configurations use the same command, arguments, and environment variable.

## Tools and capabilities

The genrelay-mcp MCP server provides four tools:

- `generate_video` submits text-to-video or image-to-video requests. It returns the clip when available, or a task ID while the job is still running.
- `generate_image` handles image generation and editing, returning the result inline.
- `check_job` checks a previously submitted video job by task ID.
- `list_models` reports the available models and what each is intended for.

Together, these tools support agent workflows that discover a suitable model, request an image or video, and poll an asynchronous video job without resubmitting it.

## Limitations and notes

Video generation can exceed the time allowed for a single MCP tool call, so receiving an in-progress task is expected behavior rather than an error. The client or agent must call `check_job` later to retrieve the result. Resubmitting an unfinished request may incur a second charge.

The README names Claude Desktop and Cursor as supported examples and states that any MCP client able to launch the stdio command can use the server. GenRelay uses prepaid USD credits on a pay-as-you-go basis. The server itself does not document local model execution; requests use the GenRelay API.

_Full upstream README: https://allmcps.com/mcp/genrelay-mcp/readme_