Local-first MCP server for constraint solving with MiniZinc and OR-Tools CP-SAT
Wellknown found it in public sources; nobody has proven control of it yet. Claiming takes one click if the repository is under your GitHub account, or a small file on your domain otherwise. Verified owners get the badge, 15-minute checks, status alerts, edits that outrank crawled data, and a ranking boost.
Agents can do it too: POST https://wellknown.network/api/v1/claims with {"agent":"openconstraint-mcp","method":"well_known_file"} — machine-readable steps at claim.json, guide at /docs/claim.
Everything here was measured by our prober or read from a registry. Nothing is self-reported.
Attributed to the source that supplied each field. Treated as claims, not facts.
# openconstraint-mcp [](https://github.com/Openconstraint/openconstraint-mcp/actions/workflows/ci.yml) [](https://pypi.org/project/openconstraint-mcp/) [OpenConstraint website](https://openconstraint.com/) · [Worked case studies](https://openconstraint.com/cases/) A local-first [Model Context Protocol](https://modelcontextprotocol.io) server for constraint programming and optimization. `openconstraint-mcp` gives an MCP client a deterministic way to compile-check and solve [MiniZinc](https://www.minizinc.org/) models on a **managed** solver runtime, exposing open-source solvers (OR-Tools CP-SAT by default, Chuffed as an optional verifier) over MCP stdio. Constraint problems — scheduling, rostering, assignment, routing, production planning, inventory — are exactly where a language model is most likely to produce an answer that looks right but is subtly infeasible. The division of labor here is **LLM proposes, server verifies**: the client's LLM drafts a MiniZinc model, and the local runtime compiles and solves it to produce a checked result. The server runs the solver; it never drafts a model of its own and never calls an LLM. Everything runs on your machine. No telemetry, no background network calls, and nothing leaves your machine unless you opt in — the only network access in the entire package is the runtime download you trigger explicitly with `install-runtime`. ## Design principles - **Local-first.** Solving, validation, and result inspection all run on your machine. There are no remote solving backends and no upload of your models or data. - **Managed runtime.** Solver execution always goes through a MiniZinc runtime this project resolves and controls, never an arbitrary `$PATH` binary — so a run does not depend on whatever MiniZinc happens to be installed on the host. - **LLM proposes, server verif…
Mapped onto the structured taxonomy from declared text and observed tool names. Confidence shown for derived entries.
Every source is kept verbatim. Field changes are logged as events.