Onomeo

What onomeo MCP server does

The onomeo MCP server gives an MCP-compatible coding agent access to models hosted by onomeo. The service exposes a rotating selection of free and paid models, including models from providers such as DeepSeek, GLM, Gemini, and Mistral. Availability can change, so the server includes a way to inspect the current model list before sending a request.

The main use case is obtaining another model’s view of a technical question. An agent can submit a proposed code change for review, ask for help with a regex, or request several independent answers to compare. The service is also suitable when the primary model needs a second opinion without leaving the current MCP client.

How it works

The server communicates with onomeo through its API and requires an onomeo API key for normal requests. list_models reports models currently served by the API, separating free models from pay-as-you-go options and showing the current free-call allowance. ask_model sends a prompt to one selected model and returns its answer. If no model is specified, auto selects a healthy free model and identifies the model that responded.

compare_models sends one prompt to between two and four models concurrently, then returns their responses together. Each model in a comparison consumes one API call. The server can be run through an MCP client using its Node.js package command, or through the packaged Claude Desktop application release.

Setup and configuration

Create an account at onomeo.com and generate an API key. A payment card is not required for free models. Node.js 20 or newer is required when running the server through npx.

For clients that use an mcpServers configuration, define the server command as npx with the GitHub repository reference and provide ONOMEO_API_KEY in the environment. Claude Code can add the same configuration through its MCP command. Claude Desktop users can instead download the .mcpb file from the latest release and enter the key when prompted.

The ONOMEO_BASE_URL variable can override the default API host. Tests normally use a local stub; setting ONOMEO_LIVE=1 with an API key enables a test containing one real auto request.

Tools and capabilities

• list_models: Shows currently available free and paid models and the live free-call limit.
• ask_model: Sends a prompt to one model, with auto available for automatic free-model selection.
• compare_models: Queries two to four models in parallel and presents their answers side by side.

Limitations and notes

Free accounts currently receive 30 calls per five-hour period, with automatic refills. A free request can contain up to 50,000 characters. Paid models, including Claude, GPT, Grok, and Gemini Pro, use pay-as-you-go billing beginning with a 5 USD top-up and do not require a subscription.

The onomeo MCP server is in public beta, so individual features may change or fail. Free models are delivered through third-party free tiers, and some may use submitted prompts for training. Review the information on each model’s page and avoid sending secrets, credentials, or private source code to free models.