Connects Claude to production optimization solvers (HiGHS, OSQP, OR-Tools) via local MCP. Exposes tools to solve LP, MIP, quadratic portfolio problems, and workforce scheduling from JSON or Excel. The check_feasibility tool computes irreducible infeasible sets and ranks constraint relaxations when no solution exists, which is often more valuable than approximate answers. Returns dual values, reduced costs, and shadow prices for sensitivity analysis. Also includes read_data_file for ingesting spreadsheets and generate_template for scaffolding problem structure. Runs synchronously but supports iterative solvers that can checkpoint and resume. Useful when you need mathematically certified solutions instead of LLM approximations for resource allocation, scheduling, or portfolio construction.
SAGE grounds AI in mathematical truth. It is a local MCP server that gives Claude Desktop — and any MCP-compatible agent — the ability to formulate, solve, and certify mathematical optimization problems using production-grade open-source solvers.
Status: v0.1.3 — Alpha · Author: Peter Pragnakar Atreides
Large Language Models are probabilistic text generators. When you ask an LLM to allocate a budget, design a schedule, optimize a route, or balance a portfolio, it generates text that resembles a solution. No simplex method runs underneath. No branch-and-bound search. No constraint check. The model cannot prove optimality, certify feasibility, or — critically — declare with certainty that no feasible solution exists.
One of the most valuable outcomes in decision-making is a mathematically certified statement of infeasibility. It tells decision-makers their goals conflict, their assumptions are inconsistent, or their constraints must be renegotiated. LLMs have no native mechanism to produce this. SAGE provides it.
SAGE introduces a hybrid intelligence architecture: LLMs handle language and ambiguity; solvers handle optimality and feasibility. Each component does what it is best suited for.
LLMs operate as single-pass inference systems — token generation stops when the response is done. Optimization solvers work differently: they are inherently iterative and stateful, designed to run for minutes, hours, or days while continuously improving. At any point they can return the best solution found so far, a bound on the optimal objective, and a certificate of optimality or infeasibility.
This "anytime" property enables SAGE to:
The result: AI shifts from immediate but approximate to sustained and mathematically grounded.
| Capability | Detail |
|---|---|
| Problem types | LP, MIP, Portfolio Optimization (QP), Workforce Scheduling |
| Solvers | HiGHS (LP/MIP), OSQP (QP) |
| File I/O | Read/write Excel (.xlsx) and CSV |
| Infeasibility | IIS detection + ranked relaxation suggestions |
| Sensitivity | Dual values, reduced costs, allowable ranges |
| Explanation | Plain-language narration of every result |
# From PyPI (once published)
pip install sage-solver-mcp
# From source (development)
git clone https://github.com/pragnakar/Project_Sage
cd sage
pip install -e sage-solver-core/
pip install -e sage-solver-mcp/
Find your config file:
~/Library/Application Support/Claude/claude_desktop_config.json%APPDATA%\Claude\claude_desktop_config.jsonAdd the SAGE server:
{
"mcpServers": {
"sage": {
"command": "uvx",
"args": ["sage-solver-mcp"]
}
}
}
What is
uvx? It is a command from the uv Python toolchain that runs a PyPI package ephemerally — no manualpip installrequired. If you have uv installed (brew install uvon macOS),uvx sage-solver-mcpfetches and runs SAGE automatically. If Claude Desktop cannot finduvxon your PATH, use the full path:"/opt/homebrew/bin/uvx"(macOS) or the output ofwhich uvx.
Restart Claude Desktop and you will see the SAGE tools in the toolbar.
Ask Claude:
"Solve this LP: maximize 3x + 5y subject to x + 2y ≤ 12, x ≤ 8, y ≤ 5, x,y ≥ 0"
Or with a file:
"Read examples/portfolio_5_assets.xlsx and solve it as a portfolio optimization"
| Tool | Description |
|---|---|
solve_optimization | Solve LP / MIP / portfolio / scheduling from JSON |
read_data_file | Read an Excel or CSV file and return a preview |
solve_from_file | Read + solve + write results in one step |
explain_solution | Narrate the most recent solve result |
check_feasibility | Check feasibility; if infeasible, compute IIS |
generate_template | Create a blank Excel template for a problem type |
suggest_relaxations | Rank constraint relaxations for the last infeasible result |
Each example shows the user prompt, which tool is called, a representative input payload, and the output SAGE returns.
User prompt: I need to figure out how many full-time and part-time employees to schedule to minimize cost. Full-time costs $200/day and covers 8 hours, part-time costs $100/day and covers 4 hours. I need at least 40 hours covered each day and at most 6 full-time staff.
Tool: solve_optimization
{
"problem_type": "lp",
"name": "staffing",
"variables": [
{"name": "ft", "lb": 0, "ub": 6},
{"name": "pt", "lb": 0}
],
"constraints": [
{"name": "coverage", "expression": {"ft": 8, "pt": 4}, "sense": ">=", "rhs": 40}
],
"objective": {"sense": "minimize", "coefficients": {"ft": 200, "pt": 100}}
}
Output: Optimal: ft=2, pt=6, cost=$1,000/day. The coverage constraint is binding. Sensitivity: each additional required hour costs $25.
User prompt: My shift schedule says workers need at least 3 people on Monday AND no more than 2 people total — is that solvable?
Tool: check_feasibility
{
"problem_type": "lp",
"name": "schedule_check",
"variables": [{"name": "workers", "lb": 0}],
"constraints": [
{"name": "min_staff", "expression": {"workers": 1}, "sense": ">=", "rhs": 3},
{"name": "max_staff", "expression": {"workers": 1}, "sense": "<=", "rhs": 2}
],
"objective": {"sense": "minimize", "coefficients": {"workers": 0}}
}
Output: INFEASIBLE. Conflicting constraints: min_staff (≥3) and max_staff (≤2) are mutually exclusive. Suggestion: relax max_staff to ≥3 (+50%) or reduce min_staff to ≤2 (−33%).
User prompt: I have a portfolio spreadsheet with expected returns and a covariance matrix. Optimize it for a target return of 8% while minimizing risk.
Tools: read_data_file → solve_from_file
read_data_file output: Detected sheets: assets (5 rows, columns: ticker, expected_return), covariance (5×5 matrix). Preview looks correct.
solve_from_file output: Optimal allocation — AAPL: 32%, MSFT: 28%, GOOGL: 18%, BND: 22%, CASH: 0%. Portfolio variance: 0.0042 (σ=6.5%). Results written to portfolio_optimized.xlsx.
User prompt: Can you create a scheduling template I can fill in? Then after I solve it, give me a detailed explanation.
Step 1 — Tool: generate_template with problem_type: "scheduling"
Output: Template written to scheduling_template.xlsx with sheets: workers (name, availability, cost), shifts (name, start, end, required_count), instructions.
Step 2 — Tool: explain_solution with detail_level: "detailed"
Output: "The optimal schedule assigns Alice and Bob to the morning shift (cost: $480) and Carlos to the evening shift (cost: $220). The evening minimum-staffing constraint has a shadow price of $45 — each additional required worker increases cost by $45. The morning capacity constraint has 1 unit of slack."
User prompt: I want to buy whole units of 3 products to maximize profit, but I can only spend $500 and store 20 cubic feet. Product A: $80, 3 ft³, $120 profit. Product B: $50, 5 ft³, $70 profit. Product C: $120, 2 ft³, $200 profit.
Tool: solve_optimization (MIP with integer variables A, B, C; budget ≤ 500; storage ≤ 20; maximize 120A + 70B + 200C)
Tool: suggest_relaxations (called automatically on infeasible sub-problem)
Output: Optimal integer solution: A=2, B=0, C=3, profit=$840. If the budget constraint is binding, suggest_relaxations ranks options: relax budget by $20 (+4%) to $520, or drop 1 unit of C and add 1 unit of A for $760 profit within the original $500 limit.
| File | Problem | Result |
|---|---|---|
examples/portfolio_5_assets.xlsx | Portfolio QP — 5 assets (equity + bonds) | Optimal allocation |
examples/nurse_scheduling.xlsx | Scheduling MIP — 8 nurses, 3 shifts, 7 days | Infeasible: IIS computed |
examples/transport_routing.xlsx | Transport LP — 3 warehouses → 5 stores | Optimal routes, $2,472 cost |
examples/blending_problem.xlsx | Blending LP — 6 ingredients, nutrient constraints | Optimal blend, $23.47/100kg |
Project_Sage/
├── sage-solver-core/ # Pure optimization engine — solver, models, fileio, explainer
│ └── sage_solver_core/
│ ├── models.py # Pydantic models (LPModel, MIPModel, PortfolioModel, SchedulingModel)
│ ├── solver.py # HiGHS + OSQP solver adapters
│ ├── builder.py # JSON → SolverInput builders
│ ├── fileio.py # Excel/CSV read/write, template generation
│ └── explainer.py# Natural language solution narration + IIS explanation
├── sage-solver-mcp/ # Local MCP server (this package — v0.1)
├── sage-solver-cloud/ # Cloud API (future — v0.2)
└── examples/ # Ready-to-use example files
Data flow:
Claude Desktop → stdio JSON-RPC → sage-solver-mcp → sage-solver-core → HiGHS/OSQP
↓
SolverResult + IIS + Sensitivity
Variables with continuous bounds, linear objective, linear constraints (<=, >=, =).
Same as LP but variables can be continuous, integer, or binary.
Markowitz mean-variance: minimize risk (quadratic) for a target return, with optional sector and weight constraints.
Assign workers to shifts over a planning horizon. Constraints: min/max workers per shift, rest periods, skill requirements.
| Phase | Focus |
|---|---|
| v0.1 (now) | LP, MIP, Portfolio QP, Scheduling — 7 MCP tools, local stdio server |
| v0.2 | sage-solver-cloud FastAPI — remote deployment, async long-running solves |
| v0.3 | Simulation — Monte Carlo, discrete-event, stochastic programming |
| v1.0 | Decision Intelligence Platform — industry templates, solver marketplace |
The long-term ambition is a planetary-scale optimization fabric: interconnected, federated models that co-optimize transportation, energy, supply chains, and infrastructure across institutions — turning SAGE from a single-user tool into shared decision infrastructure.
See CONTRIBUTING.md for dev setup, test instructions, and branch conventions.
393 tests · 0 failures · sage-solver-core 0.1.3 · sage-solver-mcp 0.1.3
MIT — Copyright (c) 2026 Peter Pragnakar Atreides
io.github.ericm1018/skillfm-llm-cost-optimizer-openai-anthropic-usage
io.github.mikerawsonnz/llm-orchestration-agent
io.github.mikerawsonnz/authenticated-llm-agent
labforgedev/copilot-memory-mcp
csoai-org/agent-prompt-injection-firewall-mcp
io.github.mikerawsonnz/authenticated-multi-llm-agent