morluto/jacobian
Pure mathematics for agents: search for examples and counterexamples, compute exactly, and independently check
Jacobian is an MCP server providing an executable vocabulary of typed mathematical operations for AI agents. math.find discovers an operation by description or name; math.run executes exactly one bounded mathematical contract and returns a typed result — exact where claimed, with approximation, incompleteness, or uncertainty made explicit (never hidden). Rather than exposing large domain solvers, Jacobian exposes mathematical primitives that agents can search for and compose into solutions. The same library is available as a CLI and native Python API alongside the MCP interface.
Install
pip install jacobianAlso available as npm: npm install jacobian. Runs as an MCP server. Configure in DSH MCP settings or add via: dsh mcp add jacobian. Server entrypoint: python -m jacobian (Python) or npx jacobian (Node).
Compatibility
Python 3.10+ or Node 18+. MCP protocol. Integrates with any MCP-capable host including DSH. MIT license.
Details
- Repo: morluto/jacobian
- Category: Agent Capabilities
- Stars: 55
- Version: Latest stable on PyPI (pip install jacobian) and npm (npm install jacobian)
- Last push: 2026-09-21
- First seen: 2026-07-23
FAQ
- What types of math operations does Jacobian support?
- Typed mathematical primitives covering algebra, calculus, linear algebra, statistics, and more — each as a bounded operation with a declared exactness contract. Use math.find to search by description.
- Is the output exact or approximate?
- Each operation declares its precision contract. Results are exact where claimed; approximation, incompleteness, or uncertainty is always made explicit in the typed result.
- How do I add Jacobian to DeepSeek Harness?
- Install with pip install jacobian or npm install jacobian, then add it as an MCP server in your DSH MCP configuration. The server exposes math.find and math.run as model-facing tools.
Alternatives
SepineTam/mcp-for-stata · anysearch-team/anysearch-dsh