OpenRouter MCP Multimodal

One MCP server gives the agent chat with hundreds of OpenRouter models, image, audio and video analysis and media generation

  • 103 ★ on the repository
  • Last commit
  • Apache-2.0
MCP server

Extended access

Why this level

  • Paid calls spend the balance of your OpenRouter key and requests and files go to external models
  • Tools read and write local files inside the allowed directories
All reasons and checks

stabgan/openrouter-mcp-multimodal

Install

Manual install

npx -y @stabgan/openrouter-mcp-multimodal

The `OPENROUTER_API_KEY` variable is required; Node.js 22 or newer.

Access and checks

How we review

Extended accessThe tool runs code or makes network calls and only reads external data.

  • Paid calls spend the balance of your OpenRouter key and requests and files go to external models
  • Tools read and write local files inside the allowed directories

This is third-party code. Review the repository files before installing.

What it does

The server connects the agent to OpenRouter and exposes 21 tools in seven groups. Chat runs through `chat_completion` with model choice, `:free` and `:online` suffixes, web search, JSON Schema structured output and async jobs. Vision and audio are covered by `analyze_image`, `generate_image`, `analyze_audio`, speech recognition and synthesis. Video has `analyze_video` and generation from text or an image with progress notifications. Separate tools give embeddings (`embed_text`), document reranking, model search and validation against the OpenRouter catalog, and a `decide` tool for yes or no decisions with a probability score. For files there are input and output path sandboxes and SSRF protection.

Who it is for. Developers who need different models and multimodal tasks from Cursor or Claude through a single OpenRouter key.

Good fit when

  • The agent must compare answers of several models on one prompt through `chat_completion`
  • Text must be pulled from a screenshot or an image described with `analyze_image`
  • A call recording needs a transcript and the project already has an OpenRouter key
  • A clip is needed from text or a frame and the wait for generation comes with progress notifications
  • Document search needs embeddings and reranking without a separate service

Not a fit when

  • Only one provider is needed directly: its own SDK is simpler and has no intermediary
  • There is no OpenRouter key and no wish to create one: the server does not run without a key
  • Data must not go to external models: requests go to OpenRouter and then to providers

Example request

Describe what is shown in screenshots/error.png, then ask two different models how to fix this error and compare the answers

MCP

Transport
stdio
Authentication
API key

Environment variables

OPENROUTER_API_KEY required, secret
OpenRouter key from openrouter.ai/keys.

Limitations

An `OPENROUTER_API_KEY` and Node.js 22 or newer are required. Free models such as `google/gemma-4-26b-a4b-it:free` cover chat and image analysis, while video and audio generation usually needs credits on the balance. Paid calls spend the user money and limits are set by OpenRouter. Payment and access to OpenRouter from Russia depend on the service terms, so check the status before installing. The `uvx` version only launches the Node package, so Node.js is still required.

How to disable. Remove the `openrouter` entry from `mcpServers` in the client config and revoke the key at openrouter.ai/keys.

FAQ

How do I connect OpenRouter MCP Multimodal to Cursor?

Set `OPENROUTER_API_KEY` and add an `openrouter` entry to `.cursor/mcp.json` with the `npx` command and the arguments `-y @stabgan/openrouter-mcp-multimodal`. A one-click install link also exists.

Can I start without paying?

Yes, free models with the `:free` suffix work for chat and image analysis. Audio and video generation usually needs credits.

Which install methods exist?

`npx`, global `npm`, Docker Hub and GHCR, Smithery and `uvx`, which launches the Node package through a thin Python wrapper.

What license applies?

Apache-2.0.

How do I choose a model?

Through the model parameter of `chat_completion`; `search_models` and `validate_model` help find and check a name in the OpenRouter catalog.

Discussion

No questions or reviews yet.

Needs a VPN

An MCP server that lets Claude Code, Codex and other agents call Gemini, GPT, Grok and local models for review, debugging and consensus

MCP serverExtended access11.8KRepository stars
Official

Official OrcaRouter MCP server: a model catalog with pricing and chat with auto-routing and a fallback chain.

MCP serverExtended access14Repository stars
Needs a VPN

A desktop app with a local MCP server that lets ChatGPT work with files, commands, skills and other MCP servers

MCP serverBroad access394Repository stars

An MCP server that inserts a human feedback step into agent tasks: a web or desktop window for text, images and prompts

MCP serverExtended access3.8KRepository stars

MCP server that gives an agent probabilities and a pick from set options, like ticket urgency, so code can branch on a number

MCP serverExtended access361Repository stars

MCP server for a shared agent room: decisions, tasks verified by another agent and a final report, joined by code.

MCP serverExtended access107Repository stars

README in short

A capability table by group, quick start with a key, install methods and configs for Cursor, VS Code, Claude Desktop and Windsurf, then security, troubleshooting and development.

README badge

Report a problem