Show HN: TabPFN MCP, gives LLMs tools for predictions on tabular data (beta)

TabPFN MCP is a Model Context Protocol server that enables LLMs and AI agents to perform tabular machine learning predictions without manual preprocessing. It provides two tools—fit_and_predict and predict—that handle missing data imputation, categorical encoding, and feature preprocessing automatically, compatible with ChatGPT, Claude, n8n, and other platforms via streamable HTTP.

Key Points

• •TabPFN MCP is a Model Context Protocol server that integrates TabPFN (a foundation model for tabular ML) with AI agents to handle structured data predictions
• •Solves the token waste problem where agents generate unreliable ML code by providing two ready-to-use tools: fit_and_predict and predict
• •TabPFN handles data preprocessing natively—no need for agents to manually impute missing values, encode categorical features, or clean messy tables
• •Compatible with major LLM platforms including ChatGPT, Claude, n8n, and others via streamable HTTP for broad integration
• •Ideal for agents working with business-critical tabular data like sales pipelines, customer records, inventory systems, and financial data
• •Currently in beta with limited access to enable close collaboration with early adopters and real-world use case validation
• •Reduces complexity and token consumption by abstracting away ML engineering tasks that agents typically struggle with

Workflow Diagram

Concepts