USolver
Provides a unified interface to numerical, logical, and symbolic solvers, leveraging MCP for problem-solving.
About
USolver is a universal solver interface designed to tackle a wide array of problems using a combination of language models and efficient solvers. It employs the "LLM sandwich" model, interpreting queries with an LLM, calling specialized solvers (like ortools, cvxpy, and z3), and then verbalizing the results. This allows for chaining solver solutions together, enabling the resolution of complex multi-step problems. It's designed to be integrated with MCP, facilitating its use with tools like Claude Desktop and Cursor.
Key Features
- 5 GitHub stars
- Supports chained solver solutions for complex problems
- Leverages LLMs for query interpretation
- Unified interface for multiple solvers (ortools, cvxpy, z3)
- Integrates with MCP for use with Claude Desktop and Cursor
Use Cases
- Solving combinatorial optimization problems
- Solving convex optimization problems
- Solving satisfiability modulo theories (SMT) problems
Sponsored