MathKernel MCP ships 160+ math tools with trust labels so models can't fake proofs
Staatsgeheim_ · reddit · 2026-09-07
The author releases MathKernel, an open-source MCP server (stdio, FastMCP 3) wrapping a Python math kernel with 160+ math tools. Core design: the model may parse and plan, but it cannot launder a symbolic result into a proof.
Key points:
- Workflow: mathcapabilities discovery → parse ASCII/LaTeX to exprid → solve/simplify/differentiate/integrate/matrices/QE/prove
- Every result carries evidencebundle, semanticstatus, and a conservative trust summary: formal / exact / symbolic / intervalcertified / numeric
- Decimals cap trust at numeric; formal certificates are refused on approximate input
- Distinct outcomes: doesnotexist, undefined, infeasible, unsupported, unknown
- Long jobs via mathjobsubmit / status / result
- Optional exact finite dynamics (Koopman, GF(2^m) closures) and portable viz/sonify artifacts that cannot raise the trust label
Python 3.11+, MIT, install via pip install 'mathkernel[mcp]'. The author stresses this is not "Claude can do math now" — most calculus remains symbolic-engine evidence, and Lean support is optional and fragmentary.
More from coding & agent
- Real-time semianalytic cloud rendering in the browser, runs even on an iPhone — NachoSoto · 2026-09-07
- Dev uses Astra to port Simpsons Hit and Run PS2 assets into a three.js web port — CtrlAltDwayne · 2026-09-07
- Goal isn't producting code but building great products, argues dev — ayushtweetshere · 2026-09-07
- Loops + graph engineering: the full guide from prompt to reliable AI systems — goyalshaliniuk · 2026-09-07
- Astra praised for proactively spawning subagents, a big force multiplier — max_paperclips · 2026-09-07
- Pinokio update dedupes files across AI apps to save terabytes of disk space — cocktailpeanut · 2026-09-07