← Back to the wire

GitHub - Staatsgeheim/MathKernel: Evidence-aware multi-engine mathematics runtime for LLMs: exact, symbolic, formal, certified-interval and numeric computation with typed MathIR, trust labels, and full provenance

AnnouncementProductSep 6, 2026

MathKernel is an evidence-aware multi-engine mathematics kernel available as a Python library and as the MCP server mathkernel-mcp, which exposes 162 tools prefixed math_ over stdio. The kernel assigns results a trust level, engine tag, and derivation trail, distinguishing exact computation, checked certificates, certified enclosures, and formal proofs. Lean 4 with Mathlib installs by default on first server start, and optional extras add GPU support via CuPy and JIT kernels via numba.

Receipt № 17741 source · awaiting confirmation ◐

Evidence

1source· awaiting independent confirmation

No score is assigned. Sources and their independence are shown in the citation chain below.

Citation chain · 1 source

01medPRIMARY
MathKernelModelmathkernel-mcpModel
Canonical: https://github.com/Staatsgeheim/MathKernel