Home / MCP / any-chat-completions-mcp MCP Server

any-chat-completions-mcp MCP Server

Provides an MCP server that relays chat queries to multiple AI providers via a unified chat completions API.

typescript

145stars

View on GitHub

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
    "mcpServers": {
        "chat_openai": {
            "command": "npx",
            "args": [
                "-y",
                "@pyroprompts/any-chat-completions-mcp"
            ],
            "env": {
                "AI_CHAT_KEY": "OPENAI_KEY",
                "AI_CHAT_NAME": "OpenAI",
                "AI_CHAT_MODEL": "gpt-4o",
                "AI_CHAT_BASE_URL": "https://api.openai.com/v1"
            }
        }
    }
}

You set up this MCP server to route chat prompts from any OpenAI‑SDK compatible API to multiple AI providers (OpenAI, PyroPrompts, Perplexity, and others) using a single, unified chat completion interface. This enables you to swap providers or run multiple providers in parallel without changing your client integrations.

How to use

Start by configuring MCP entries for each provider you want to use. Each entry runs as a local, stdio MCP server that you connect to from your client. You can have a single provider or multiple providers, and you can switch between them by selecting the corresponding MCP server in your client configuration. All providers appear as separate tools within your chat environment, so you can address each one by name and route questions accordingly.

How to install

Prerequisites: you need Node.js and npm installed on your system.

Install dependencies and build the server if you are using a local clone. Run the following in your terminal where the project resides.

npm install

npm run build

To develop with auto-rebuild, you can use the watch script during development.

npm run watch

Configure your client to load the MCP entries that point to the local stdio servers described below.

Provider configurations (stdio MCP servers)

You can run multiple providers by creating separate MCP server entries. Each entry uses stdio and points to the built index of the MCP server, with environment variables identifying the provider.

{
  "mcpServers": {
    "chat-openai": {
      "command": "npx",
      "args": [
        "-y",
        "@pyroprompts/any-chat-completions-mcp"
      ],
      "env": {
        "AI_CHAT_KEY": "OPENAI_KEY",
        "AI_CHAT_NAME": "OpenAI",
        "AI_CHAT_MODEL": "gpt-4o",
        "AI_CHAT_BASE_URL": "https://api.openai.com/v1"
      }
    },
    "chat-pyroprompts": {
      "command": "node",
      "args": [
        "/path/to/any-chat-completions-mcp/build/index.js"
      ],
      "env": {
        "AI_CHAT_KEY": "PYROPROMPTS_KEY",
        "AI_CHAT_NAME": "PyroPrompts",
        "AI_CHAT_MODEL": "ash",
        "AI_CHAT_BASE_URL": "https://api.pyroprompts.com/openaiv1"
      }
    },
    "chat-perplexity": {
      "command": "node",
      "args": [
        "/path/to/any-chat-completions-mcp/build/index.js"
      ],
      "env": {
        "AI_CHAT_KEY": "PERPLEXITY_KEY",
        "AI_CHAT_NAME": "Perplexity",
        "AI_CHAT_MODEL": "sonar",
        "AI_CHAT_BASE_URL": "https://api.perplexity.ai"
      }
    }
  }
}

Security and usage notes

Keep API keys and base URLs secure. Do not expose keys in client configurations or public repositories. Use environment isolation where possible and rotate keys periodically.

Each provider’s configuration is isolated by its own MCP entry. If you need to disable a provider, remove or comment out the corresponding MCP server entry and restart the client.

Debugging and troubleshooting

MCP servers communicate over stdio, which can be tricky to debug. If you need debugging tools, run the inspector to access debugging utilities in your browser or terminal.

npm run inspector

Available tools

chat

Relays a user question to a configured AI chat provider and returns the provider’s response, enabling multi-provider chat through a single interface.