morluto/jacobian

What it solves

AI agents often struggle with complex mathematical reasoning because they lack reliable, precise tools. Jacobian provides a searchable, executable library of mathematical operations that allows agents to perform exact calculations rather than relying on probabilistic guesses or pre-built, rigid workflows.

How it works

Jacobian acts as an MCP (Model Context Protocol) server. It exposes a set of "typed" mathematical operations that agents can discover using math.find and execute using math.run. Each operation is designed to be semantically narrow and self-contained, returning exact results. Agents are responsible for choosing which operations to call, how to combine the results, and when to stop, giving them the flexibility to solve problems that go beyond any single pre-defined tool.

Who it’s for

This project is for developers building AI agents who need to perform trustworthy, precise mathematical operations—such as linear algebra, graph theory, or SAT/SMT solving—within their agentic workflows.

Highlights

• Provides a searchable, typed vocabulary of mathematical primitives.
• Supports complex domains including polynomial algebra, graph theory, and bounded SAT/SMT solving.
• Operates as an MCP server for easy integration with existing AI agent frameworks.
• Offers a consistent Python API and CLI for local terminal workflows.
• Focuses on semantic atomicity, ensuring each tool performs one stable, reusable mathematical move.