{"id":"6368105c-7686-4273-a3dd-bc7eef25f023","slug":"reasonforge","kind":"mcp","name":"ReasonForge","summary":"Modular Mathematics Toolkit MCP Server: 111 tools that give LLMs exact math instead of hallucinated guesses. Routes calculations through SymPy/NumPy/SciPy for provably correct results. 7 modular packa","owner":"foxintheloop","repoUrl":"https://github.com/foxintheloop/ReasonForge","license":"MIT","licenseTier":"permissive","trust":"community","stars":0,"match":null,"fewStars":true,"format":"mcp","createdAt":"2025-11-28T04:09:46.000Z","addedAt":"2026-10-21T05:43:56.266Z","logoUrl":"https://avatars.githubusercontent.com/u/8127922?u=0cf65bedfb6a80536a4900cbf7410fdc89042792&v=4","linkOnly":false,"collection":null,"skillPath":null,"categories":["cloud-infra","memory-knowledge"],"tags":["claude","llm","mathematics","numpy","python","scipy","symbolic-computation","sympy"],"runtime":"binary","local":true,"cloudCapable":false,"cloudCapability":"unknown","cloudReadyOnPlatform":false,"needsApiKey":false,"reviewed":false,"syncedAt":"2026-10-23T21:06:13.579Z","sample":false,"url":"https://conduits.dev/tools/reasonforge","language":"HTML","transports":["stdio"],"install":{},"requirements":{"env":[],"upstreamMayCharge":false},"tools":null,"readme":"# ReasonForge\n\n**Exact math for LLMs. No hallucinations.**\n\nReasonForge is a modular mathematics toolkit that integrates with Claude and other LLMs via MCP (Model Context Protocol). Instead of letting the LLM guess at calculations, ReasonForge routes mathematical operations through SymPy, NumPy, and SciPy — returning provably correct results.\n\n111 tools. 7 domain-specific servers. Zero probabilistic math.\n\n## The Problem\n\nAsk an LLM to solve a differential equation, simplify a complex expression, or find a sequence pattern, and you'll get plausible-looking answers that are wrong 15-30% of the time. For research, finance, or engineering applications, \"usually right\" isn't good enough.\n\n## The Solution\n\nReasonForge gives your LLM access to a real computation engine. The LLM handles natural language understanding and problem decomposition; ReasonForge handles the actual math.\n\n```\nUser: \"Find the pattern in [1, 4, 9, 16, 25] and give me the equation\"\n\nClaude + ReasonForge:\n  → Calls find_sequence_pattern tool\n  → Returns: n² (exact symbolic formula, not a guess)\n```\n\n![ReasonForge Demo in Claude Desktop](assets/find-the-pattern.png)\n\n## Quick Start\n\n```bash\npip install reasonforge-logic\n```\n\nOr install the complete ecosystem:\n\n```bash\npip install reasonforge[all]\n```\n\nOr install from source:\n\n```bash\ngit clone https://github.com/foxintheloop/ReasonForge.git\ncd ReasonForge\npip install -e packages/reasonforge -e packages/reasonforge-logic\n```\n\nAdd to `claude_desktop_config.json`:\n\n```json\n{\n  \"mcpServers\": {\n    \"reasonforge-logic\": {\n      \"command\": \"python\",\n      \"args\": [\"-m\", \"reasonforge_logic\"]\n    }\n  }\n}\n```\n\nDone. Claude now has access to 13 symbolic reasoning tools.\n\n## Example Queries\n\n- *\"Convert (A OR B) AND C to conjunctive normal form\"*\n- *\"Prove that a + b = b + a using commutativity\"*\n- *\"Solve the system: 2x + y = 5, x - y = 1\"*\n- *\"Find the derivative of x³ sin(x)\"*\n\n## Architecture\n\nInstall only what you need:\n\n```\nReasonForge Ecosystem\n│\n├── reasonforge-core          # Pure Python symbolic engine (no MCP)\n│\n├── reasonforge-logic         # 13 tools ✅ Production Ready\n│   └── Symbolic reasoning, formal logic, knowledge systems\n│\n├── reasonforge-expressions   # 16 tools (Beta)\n│   └── Variable management, expression operations, basic calculus\n│\n├── reasonforge-algebra       # 18 tools (Beta)\n│   └── Equation solving, matrices, optimization\n│\n├── reasonforge-analysis      # 17 tools (Beta)\n│   └── Differential equations, transforms, signal processing\n│\n├── reasonforge-geometry      # 15 tools (Beta)\n│   └── Vector/tensor calculus, coordinate systems\n│\n├── reasonforge-statistics    # 16 tools (Beta)\n│   └── Probability, distributions, hypothesis testing\n│\n└── reasonforge-physics       # 16 tools (Beta)\n    └── Classical mechanics, quantum states, electromagnetism\n```\n\n**Why modular?** A monolithic server loads ~8,000 tokens of tool definitions into context. A single package loads ~800-1,200 — an 85-90% reduction.\n\n## Package Status\n\n| Package | Tools | Status | Use Case |\n|---------|-------|--------|----------|\n| reasonforge-logic | 13 | ✅ Ready | Formal proofs, boolean logic, pattern recognition |\n| reasonforge-expressions | 16 | Beta | Expression simplification, substitution, basic calculus |\n| reasonforge-algebra | 18 | Beta | Linear systems, polynomial roots, matrix operations |\n| reasonforge-analysis | 17 | Beta | ODEs, PDEs, Laplace/Fourier transforms |\n| reasonforge-geometry | 15 | Beta | Vectors, tensors, coordinate transforms |\n| reasonforge-statistics | 16 | Beta | Distributions, inference, regression |\n| reasonforge-physics | 16 | Beta | Mechanics, E&M, quantum computing |\n\n> **Note:** Only `reasonforge-logic` is fully tested for production. Other packages are functional but may have edge cases.\n\n## Design Goals\n\n| Goal | Status |\n|------|--------|\n| Exact symbolic computation | ✅ Implemented via SymPy |\n| Deterministic results | ✅ Achieved for supported operations |\n| Reproducibility | ✅ Same input → same output |\n| Full domain coverage | In progress |\n\n## Documentation\n\n- [Usage Examples](USAGE_EXAMPLES.md) — 110 examples across all packages\n- Individual package READMEs in `/packages/`\n\n## Contributing\n\n```bash\n# Run all tests\npytest packages/\n\n# Run specific package tests\npytest packages/reasonforge-logic/tests/ -v\n```\n\nPRs welcome for new tools, packages, or docs.\n\n## License\n\nMIT — see [LICENSE](LICENSE)\n\n## Citation\n\n```bibtex\n@software{reasonforge,\n  title = {ReasonForge: Mathematics Toolkit for LLM Integration},\n  author = {Derek Fox},\n  year = {2025},\n  url = {https://github.com/foxintheloop/ReasonForge}\n}\n```\n\n---\n\nBuilt on [SymPy](https://www.sympy.org/), [NumPy](https://numpy.org/), [SciPy](https://scipy.org/), and the [Model Context Protocol](https://modelcontextprotocol.io/).\n"}