unifai

MCP server from unifai-network/unifai-mcp-server

Installation

Run this command in your terminal to add the MCP server to Claude Code.

Run in terminal:

Command

claude mcp add --transport stdio unifai-network-unifai-mcp-server node path/to/server.js \
  --env UNIFAI_API_KEY="your-unifai-api-key" \
  --env UNIFAI_API_BASE_URL="https://api.unifai.example.com"

How to use

The UnifAI MCP Server exposes the tools from the UnifAI SDKs so MCP clients can discover and execute UnifAI capabilities in a consistent way. This server acts as a bridge that presents the SDK-provided tools (such as data processing, reasoning, and inference utilities) as MCP commands that clients can invoke. You can connect your MCP client to the server, browse the available tools, and call them with structured inputs and receive structured outputs. Since UnifAI provides both Python and TypeScript SDKs, the MCP server is designed to work with either language runtime via the same MCP protocol.

To use the server, point your MCP client to the server endpoint, authenticate with the API key if required, and request the list of available tools. Then you can execute a tool by sending the appropriate input payload as defined by the tool’s schema. The server will return results or error messages that your client can handle, allowing you to chain tools for complex workflows. If you’re using the TypeScript SDK, tool invocations can leverage the same MCP-compatible interface exposed by the server; if you’re using the Python SDK, you can mirror the same tool usage patterns through the Python client while the server handles cross-language tool invocation.

How to install

Prerequisites:

Node.js (v14+ recommended) or a compatible runtime
npm or yarn for package installation
Access to the UnifAI API if your tools require authentication

Install locally (Node.js runtime):

Clone the repository or download the release package:

git clone https://github.com/unifai-network/unifai-network-unifai-mcp-server.git cd unifai-network-unifai-mcp-server
Install dependencies:

npm install
Configure environment variables (example):

UNIFAI_API_KEY=your-unifai-api-key UNIFAI_API_BASE_URL=https://api.unifai.example.com
Start the MCP server:

npm run start

If you prefer running via a direct Node.js invocation, ensure you point to your built server entry (for example, node dist/server.js) after building the project.

Note: If your deployment uses a Python runtime or a Docker image, follow the equivalent steps for those runtimes as provided by the UnifAI SDKs and your deployment environment.

Additional notes

Tips and common issues:

Ensure your UNIFAI_API_KEY is valid and has access to the required tools.
If you see tool-not-found errors, verify that the server’s tool registry is synchronized with the UnifAI SDK version you intend to use.
For production deployments, consider securing the API key with a secret manager and restricting network access to trusted clients.
The MCP server is designed to be language-agnostic; if you switch between Python and TypeScript SDKs, simply point your MCP clients to the same MCP server endpoint.
If you need to tune performance, review the server’s tool invocation timeout settings and adjust the process concurrency as appropriate.