# stabgan/openrouter-mcp-multimodal [Health: Active]

**Category:** 🎥 Multimedia Process  
**Repository:** https://github.com/stabgan/openrouter-mcp-multimodal  
**GitHub Stars:** 101  
**npm Downloads (last month):** 1576  
**Views:** 8  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/stabgan-openrouter-mcp-multimodal

## Description
All-in-one multimodal MCP for 300+ OpenRouter models: text chat, image / audio / video analysis, and image / audio / video generation (Veo 3.1, Sora 2 Pro, Seedance, Wan). Structured meta.code error taxonomy, IPv4+IPv6 SSRF guards, path-sandbox for disk writes, retry-after-aware backoff, multi-arch Docker.

## Tools
Capabilities this server exposes over MCP:

- **chat_completion** — Text chat, web search, provider routing, caching, reasoning, structured output
- **start_chat_completion** — Async background job for long-running reasoning models
- **get_chat_completion_status** — Poll / retrieve async completion results
- **analyze_image** — Vision — local path, URL, or data URL + `question
- **analyze_audio** — Transcribe / analyze audio files
- **analyze_video** — Describe / Q&A over video files
- **generate_image** — Text-to-image via chat completions with reference images
- **generate_image_dedicated** — Text-to-image via dedicated `/api/v1/images` (resolution, quality, format)
- **generate_audio** — Text-to-speech / music via chat completions
- **text_to_speech** — Dedicated TTS (`/api/v1/audio/speech`) — Deepgram, ElevenLabs, OpenAI; voices, speed, mp3/pcm
- **speech_to_text** — Dedicated STT (`/api/v1/audio/transcriptions`) — Whisper, GPT-4o Transcribe, Voxtral, ElevenLabs Scribe
- **generate_video** — Text-to-video (async, resumable)
- **generate_video_from_image** — Image-to-video (narrower schema)
- **get_video_status** — Poll / resume video jobs
- **search_models** — Paginated model catalog search
- **get_model_info** — Pricing, context, modalities
- **validate_model** — Cheap model ID existence check
- **rerank_documents** — Relevance ranking for RAG
- **embed_text** — Vector embeddings (37+ models: text-embedding-3-small/large, Nomic, etc.)
- **decide** — Structured decisions with calibrated probabilities (System One models)
- **health_check** — API key + reachability probe

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

```json
"mcpServers": {
  "openrouter-mcp-multimodal": {
    "command": "npx",
    "args": ["-y","@stabgan/openrouter-mcp-multimodal"],
    "env": {
      "OPENROUTER_API_KEY": ""
    }
  }
}
```

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

## Documentation

## What stabgan/openrouter-mcp-multimodal MCP server does

OpenRouter MCP Multimodal gives MCP clients a single stdio connection to OpenRouter’s model catalog and API. Its 21 tools cover conversational completion, image understanding and creation, audio transcription and synthesis, video analysis and generation, embeddings, document reranking, model lookup, structured decisions, and service checks.

The chat tools support ordinary and reasoning-oriented requests, provider routing, web search, response caching, structured JSON output, and model suffixes such as `:free`, `:online`, `:nitro`, `:floor`, and `:exacto`. Vision tools accept local paths, URLs, or data URLs for image analysis. Audio and video tools handle file analysis, while generation tools cover images, speech, music, and video.

## How it works

The server forwards requests to OpenRouter, which provides access to more than 300 models. Use `chat_completion` for a direct result, or start a background request with `start_chat_completion` and retrieve it using `get_chat_completion_status`. Video generation follows a similar asynchronous pattern with `generate_video`, `generate_video_from_image`, and `get_video_status`.

Dedicated media endpoints are available where OpenRouter exposes them. `generate_image_dedicated` supports image resolution, quality, and format controls. `text_to_speech` uses the dedicated speech endpoint with provider options including Deepgram, ElevenLabs, and OpenAI, while `speech_to_text` supports transcription providers such as Whisper, GPT-4o Transcribe, Voxtral, and ElevenLabs Scribe.

## Setup and configuration

The server runs on Node.js 22 or newer. Set an OpenRouter API key and launch the package with npx:

```bash
export OPENROUTER_API_KEY=sk-or-v1-...
npx -y @stabgan/openrouter-mcp-multimodal
```

Add the resulting stdio server to an MCP client’s `mcpServers` configuration. The documented clients include Cursor, Claude Desktop, VS Code, Windsurf, and Cline. Cursor supports project or user configuration; VS Code supports workspace or user configuration; Claude Desktop uses its platform-specific application configuration file.

The package is distributed through npm, with Python and Docker package listings also available. Free OpenRouter models can be used for chat and vision, while audio and video generation commonly require account credits.

## Tools and capabilities

- Run synchronous or asynchronous chat completions with reasoning, provider routing, web search, caching, and structured output.
- Inspect images, audio, and video, including image questions and audio transcription.
- Create images, speech, music, and videos, including image-to-video jobs.
- Search, inspect, and validate OpenRouter model identifiers and retrieve pricing, context, and modality details.
- Generate embeddings with more than 37 embedding models and rerank documents for retrieval workflows.
- Produce structured decisions with calibrated probabilities and check API reachability with `health_check`.

## Limitations and notes

Media capabilities depend on the selected OpenRouter model and provider. Long-running reasoning and video requests require asynchronous polling or resumption. Video and audio generation may require credits even when free models are available for chat and vision. Local file operations are constrained by path sandboxes, and remote resource access includes SSRF protections. Errors use structured metadata including a `meta.code` classification.

_Full upstream README: https://allmcps.com/mcp/stabgan-openrouter-mcp-multimodal/readme_