From the rules to a checked solution

Your coding agent proposes a constraint model from a problem stated in plain words. The OpenConstraint server validates it and runs it on your machine. Four worked case studies show the model, the run, and a solution an independent checker verifies.

DecisionsSolver searchFinish timeRuled outRuled outValidValidOptimalValid
Case studies

Worked constraint problems

Available now

A local-⁠first MCP server

Use the open-source openconstraint-mcp server with a Model Context Protocol (MCP) client such as Claude Code or Codex to develop, check, and solve constraint models with MiniZinc and OR-Tools CP-SAT. It runs on your machine and collects no telemetry.

  1. The client's language model proposes a constraint model.
  2. The server validates the model and its data, then runs it.
  3. The solver searches for a solution.
  4. A separate checker, when you supply one, verifies the result.

Have a constraint problem?

Describe the decisions to make, the rules a valid solution must satisfy, and any objective you want to optimize. Questions about modeling, the case studies, and OpenConstraint are welcome. The issue tracker is public, so leave out confidential and personal data.