Unichat MCP Server in Python

What is Unichat MCP Server in Python?

Unichat MCP Server in Python is an MCP server that sends requests to OpenAI, Anthropic, and OpenAI-compatible providers (such as MistralAI, xAI, Google AI, DeepSeek, Alibaba, or Inception) using the Model Context Protocol. It uses one tool and several predefined prompts, and requires a vendor API key and a selected model. It is designed for developers who want to integrate AI model chat into MCP‑based applications.

How to use Unichat MCP Server in Python?

Configure the server in your MCP client (e.g., Claude Desktop) by setting the environment variables UNICHAT_MODEL, UNICHAT_API_KEY, and optionally UNICHAT_BASE_URL. Run the server via uvx unichat-mcp-server (published) or locally with uv run unichat-mcp-server. The server exposes a tool called unichat and prompts such as code_review, document_code, explain_code, and code_rework. Installation via Smithery is also supported: npx -y @smithery/cli install unichat-mcp-server --client claude.

Key features of Unichat MCP Server in Python

• Sends requests to OpenAI, Anthropic, and OpenAI‑compatible providers.
• Supports custom API endpoints via UNICHAT_BASE_URL.
• Includes one tool: unichat (requires messages argument).
• Provides four code‑oriented prompts: review, document, explain, rework.
• Works with any model supported by the Unichat library.
• Deployable via Smithery or as a published PyPI package.

Use cases of Unichat MCP Server in Python

• Integrate AI chat into an MCP‑enabled IDE or editor for code review.
• Automate code documentation generation inside a development workflow.
• Explain complex code snippets to a developer during a review session.
• Apply requested changes to code by using the code_rework prompt.

FAQ from Unichat MCP Server in Python

Which providers are supported?

OpenAI, Anthropic, and any OpenAI‑compatible provider such as MistralAI, xAI, Google AI, DeepSeek, Alibaba, or Inception.

What environment variables are required?

You must set UNICHAT_MODEL (the model name from the supported list) and UNICHAT_API_KEY (your vendor API key). For custom endpoints, also set UNICHAT_BASE_URL to the provider’s compatible endpoint.

How do I install the server?

Use uvx unichat-mcp-server for published packages, or clone the repository and run with uv run unichat-mcp-server. It can also be installed automatically via Smithery.

What tools and prompts does it provide?

One tool (unichat) that takes a messages string and returns a response. Four prompts: code_review, document_code, explain_code, and code_rework – each requiring a code string, with code_rework also accepting an optional changes string.