SymPy MCP Server

Language models are poor at symbol pushing — they invent constants, permute terms and quietly lose a sign — so the fix is to stop asking them and orchestrate a computer algebra kernel instead. Variables are introduced with explicit assumptions and kept in server-side state, expressions are parsed against those variables, and the transforms happen inside SymPy: algebraic and linear and nonlinear systems, differentiation and integration, vector calculus, ordinary and partial differential equations. Results can be printed back as LaTeX with the variable assumptions attached, which is what keeps a multi-step derivation auditable. A relativity extra adds predefined spacetime metrics such as Schwarzschild and Kerr for tensor calculations.