LabscoMCP early access ↗
novvooMCP SERVER · OFFICIAL PROJECT

Gurddy MCP

Hand over a constraint or optimisation problem — a linear program, a colouring, a portfolio, a facility siting — and get the solved answer back rather than code to run yourself.

by novvoo ↗Physical Sciences & Scientific Computing✓ Official

LABSCO SUMMARY

One tool, `install`, is the outlier: it changes the machine rather than computing an answer, adding or upgrading the gurddy package.

Everything else takes a model and returns a solution — solve_lp through PuLP, solve_scipy_portfolio_optimization and solve_scipy_facility_location through SciPy, the colouring and N-Queens solvers as constraint problems — so the blast radius is a computation, with that single exception. The range runs from textbook set pieces (solve_sudoku, solve_24_point_game, solve_chicken_rabbit_problem) to models you would put in front of a real decision (solve_production_planning with sensitivity_analysis, solve_scipy_facility_location with fixed_cost and max_facilities), which makes run_example the quickest way to see which end a given solver sits at. Input shape is uneven across the list: solve_lp takes a single problem argument where the SciPy and planning solvers take the model broken into named parts.

What it is

A solver server over the gurddy package: linear and mixed-integer programs through PuLP, constraint problems such as N-Queens, Sudoku and graph or map colouring, minimax for games and for decisions under uncertainty, and SciPy-backed portfolio, fitting and facility-location models.

What you get

  • Linear and mixed-integer programming: solve_lp takes a problem and solves it as an LP or a MIP through PuLP.
  • A planning model with its sensitivity: solve_production_planning takes capacities, consumption and profits, an integer switch, and an optional sensitivity_analysis.
  • Constraint problems solved directly rather than described: solve_n_queens takes n, solve_sudoku takes a 9x9 puzzle, solve_graph_coloring takes edges with num_vertices and max_colors, and solve_map_coloring takes regions with their adjacencies.
  • Decisions against an adversary or against uncertainty: solve_minimax_game solves a two-player zero-sum game from a payoff_matrix for a given player, and solve_minimax_decision takes decision_vars, scenarios, an objective and a budget for robust optimisation.
  • SciPy-backed numerical models: solve_scipy_portfolio_optimization takes expected_returns, a covariance_matrix and risk_tolerance; solve_scipy_statistical_fitting estimates parameters for a distribution from data; solve_scipy_facility_location combines CSP and SciPy over customer_locations, customer_demands, facility_locations, fixed_cost and max_facilities.
  • Worked examples on demand: run_example runs one of lp, csp, n_queens, graph_coloring, map_coloring, scheduling, logic_puzzles, optimized_csp, optimized_lp, minimax, scipy_optimization or classic_problems.
  • Word-problem shortcuts: solve_24_point_game finds an arithmetic expression over four numbers, and solve_chicken_rabbit_problem solves from total_heads and total_legs.

Requirements

No credentials are configured. The solvers run on the gurddy package, which install is the tool for putting in place or upgrading, and info reports on.

Setup effort

One command — pip install gurddy_mcp

See installation options