Any Chat Completions MCP

What pyroprompts/any-chat-completions-mcp MCP server does

The pyroprompts/any-chat-completions-mcp MCP server adds a chat tool to an MCP client. That tool sends a question to a configured AI chat provider through an OpenAI SDK-compatible Chat Completions API and relays the response back to the client.

The project is implemented in TypeScript. Its provider examples include OpenAI, Perplexity, Groq, xAI, and PyroPrompts, but the endpoint must follow the expected OpenAI-compatible chat API pattern. The server does not provide separate tools for each provider. Instead, each configured instance represents one provider and exposes the same chat capability.

How it works

At startup, the process reads four environment variables: an API credential, a display name, a model identifier, and the provider's base URL. The chat tool then uses those values when sending requests to the selected endpoint.

You can register the same pyroprompts/any-chat-completions-mcp MCP server more than once in an MCP client. Assigning different environment values to each registration creates separate chat tools, such as one for OpenAI and another for Perplexity. This lets the client present multiple configured providers at the same time without changing the server code.

The README documents stdio configuration for Claude Desktop and LibreChat. Claude Desktop launches the package with npx, while a locally cloned checkout can be started from its built JavaScript entry point.

Setup and configuration

The package can be used through the npm package @pyroprompts/any-chat-completions-mcp, or built from a cloned repository. A source checkout requires installing dependencies and running the build script. The project also includes a watch command for development rebuilds.

A typical client configuration supplies these values:

• AI_CHAT_KEY: the API key for the selected provider.
• AI_CHAT_NAME: the provider name shown to the client.
• AI_CHAT_MODEL: the model to request.
• AI_CHAT_BASE_URL: the compatible API's base URL.

For example, an OpenAI configuration uses the OpenAI API base URL and a model such as gpt-4o; a Perplexity configuration can use its API URL and the sonar model. The README's values are examples and must be replaced with credentials and models valid for the provider being used.

Tools and capabilities

The server exposes one tool:

• chat: sends a question to the configured chat-completions provider.

Its main capability is provider selection through configuration rather than through different tool implementations. Multiple instances can provide access to several OpenAI-compatible APIs in the same client. The documented client setups include Claude Desktop and LibreChat. The repository also provides an npm script that starts MCP Inspector for stdio debugging.

Limitations and notes

This project depends on the target provider supporting the expected OpenAI-compatible Chat Completions interface. The server does not document provider-specific features beyond forwarding a question to the configured service, so compatibility may depend on the endpoint, model, and API implementation.

Each configured instance requires its own environment values. API credentials are supplied by the user; the README does not describe a built-in account, credential store, or hosted endpoint. The pyroprompts/any-chat-completions-mcp MCP server communicates over stdio in the documented local-client setups, so the client must be able to launch a local process.