English · 简体中文

Jacobian

Atomic mathematics for agents: discover one typed operation, run it, and compose its bounded result.

Jacobian is an MCP server that gives AI agents atomic, composable tools for higher mathematics. It exposes two tools: math.find searches an immutable library of typed mathematical operations, and math.run executes exactly one of them and returns its concrete typed result. Each operation is one bounded, exact computation: a typed request in and a typed mathematical value out. The same mathematical library is also available through a CLI and native Python API.

Jacobian's hypothesis is that mathematical reasoning benefits from an executable vocabulary of small, exact operations. Rather than exposing large domain solvers or precomposed workflows, Jacobian exposes mathematical primitives that agents can search for and compose into solutions beyond what any individual operation was designed to solve. The library supplies trustworthy mathematical moves; the reasoning model decides which moves to make, how to combine their results, and when to stop. Keeping the operations small and domain-owned preserves that search space instead of baking one proof strategy or workflow into the tools themselves.

Quickstart

Run the canonical Python MCP command without installing Jacobian globally:

uvx --from jacobian jacobian-mcp

Where an MCP host requires an npm command, the npm package is a deterministic carrier for that same command:

npx jacobian mcp

For a persistent installation:

python -m pip install jacobian
jacobian-mcp

That package includes Jacobian's exact maintained Python backend stack: SymPy, NetworkX, Z3, and Python-FLINT. A normal Python or npm installation therefore exposes the same built-in Python-backed operation portfolio. The tested binary-install contract is CPython 3.12 or 3.13 on glibc Linux x86-64; the release gate installs the built wheel and starts Jacobian on both Python versions. Other systems may have compatible upstream wheels, but are not part of the tested release contract yet. In particular, Alpine/musl cannot install the complete mandatory stack from PyPI.

The Python distribution contains the mathematical kernel, CLI, and MCP server. The npm package deterministically maps its exact package version to the corresponding uvx invocation.

Compute one bounded result

An ordinary operation returns mathematics first. For example, matrix.determinant.compute accepts one exact rational matrix and returns its determinant directly. Callers compose results by passing their typed values to a subsequent operation.

Available mathematics

The built-in portfolio covers work in:

• polynomial maps and polynomial algebra;
• exact linear algebra;
• graphs, paths, colorings, and isomorphism;
• bounded SAT and SMT solving;
• finite algebra, probability, geometry, and topology; and
• Lean source elaboration.

SAT and SMT operations use the maintained Z3 Python binding directly. The optional lean.check operation runs one bounded source snippet in the fixed Lean service environment, using a request-scoped temporary directory and returning typed diagnostics. Use math.find to search for an operation, browse an unfamiliar domain, and inspect one operation before calling math.run once.

See the domain operation library for the maintained operation portfolio and backend requirements.

Status

Jacobian 0.12.0 is pre-stable. Its published package and operation contracts describe the supported surface; experimental operation contracts may change between releases.

Documentation

• Documentation home: tutorials, how-to guides, reference, and explanations
• Architecture: runtime structure and trust boundaries
• Product model: operation contracts, ownership, and project boundaries
• Tool reference: MCP resources and invocation contracts
• Backend requirements: maintained Python backends and optional Lean
• Remote deployment: HTTP deployment and authentication

Contributing

Jacobian uses Python 3.12, uv, and a small Makefile:

make setup
make test-math
make check

Read CONTRIBUTING.md before changing code. It documents focused test commands, verification rules, documentation placement, and pull-request expectations.

License

MIT