rusolve
Agent playbooks for the rusolve optimization solver
Safety scan
A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.
Badge for your README
Markdown
[](/listings/creait-nl--rusolve)HTML
<a href="/listings/creait-nl--rusolve"><img src="/api/listings/creait-nl--rusolve/badge.svg" alt="findsafeskills safety rating" /></a>Install
Install instructions from the README
As written by the repo's author. findsafeskills has not verified these steps.
From source (works today; needs a Rust toolchain):
git clone https://github.com/CREAIT-nl/rusolve && cd rusolve cargo build --release -p rusolve-cli # the solver binary: target/release/rusolve-cli
Python package (3.11 or newer; the wheel bundles the CLI and the MCP server):
pip install maturin && tools/build_wheel.sh # writes the wheel to dist-wheels/ pip install dist-wheels/rusolve-*.whl
After the first release this becomes `pip install rusolve`, or `uvx --from rusolve rusolve mcp` to run the MCP server without installing.
The CLI is called `rusolve-cli` when built with cargo and `rusolve` when installed from the wheel. The examples below use `rusolve`.
What was scanned
The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.
.claude-plugin/marketplace.json · Manifest · 14 lines
- {
- "name": "rusolve",
- "owner": {
- "name": "rusolve contributors"
- },
- "plugins": [
- {
- "name": "rusolve-skills",
- "source": "./",
- "description": "Playbook skills for formulating, solving, debugging and tuning models with rusolve"
- }
- ],
- "description": "Agent playbooks for the rusolve optimization solver"
- }
README.md · README · 122 lines
- # rusolve
- An LP and MIP solver in Rust, built to be used by people and by AI agents.
- rusolve starts from a faithful Rust port of [HiGHS](https://highs.dev): its dual simplex,
- interior point, presolve and branch-and-cut are translated function by function and checked
- against HiGHS itself. On top of that base it adds techniques from the optimization literature
- and from other permissively licensed solvers (SCIP, Clarabel, and others, credited in
- [`THIRD_PARTY_NOTICES.md`](THIRD_PARTY_NOTICES.md)). It is also designed for AI agents: every
- command can return JSON, every error has a code and an explanation, and an MCP server is built
- in.
- **Status:** early (0.1.0), not yet on PyPI. LP and MIP are the core. Convex quadratic
- objectives and semi-continuous variables are supported. Interfaces may still change.
- ## Install
- From source (works today; needs a Rust toolchain):
- git clone https://github.com/CREAIT-nl/rusolve && cd rusolve
- cargo build --release -p rusolve-cli # the solver binary: target/release/rusolve-cli
- Python package (3.11 or newer; the wheel bundles the CLI and the MCP server):
- pip install maturin && tools/build_wheel.sh # writes the wheel to dist-wheels/
- pip install dist-wheels/rusolve-*.whl
- After the first release this becomes `pip install rusolve`, or `uvx --from rusolve rusolve mcp`
- to run the MCP server without installing.
- The CLI is called `rusolve-cli` when built with cargo and `rusolve` when installed from the
- wheel. The examples below use `rusolve`.
- ## Quick start
- Python:
- ```python
- import rusolve as rs
- m = rs.Model()
- x = m.add_vars(3, binary=True)
- m.add_constr(x[0] + x[1] + x[2] <= 2)
- m.maximize(5 * x[0] + 4 * x[1] + 3 * x[2])
- sol = m.solve(time_limit=60)
- print(sol.status, sol.objective) # Optimal 9.0
- ```
- More in [`crates/rusolve-py/README.md`](crates/rusolve-py/README.md): bulk data and sparse
- matrices, warm re-solves after changes, callbacks, and SciPy- and highspy-compatible layers.
- Command line (MPS and LP files, optionally compressed):
- rusolve mip model.mps --format json --solution # solve, result as one JSON object
- rusolve explain model.mps # why is it infeasible? (an IIS, by name)
- rusolve feasrelax model.mps --mode min-count # the cheapest repair
- rusolve check model.mps solution.json # which constraints a solution violates
- rusolve lint model.mps # big-M values, scaling, modelling smells
- rusolve options # every option, its range and default
- `rusolve --help` lists all commands.
- ## Diagnostics
- Problems with a model or a command come back as coded diagnostics, in the style of `rustc`:
- warning[RS2001]: big-m: row 'c1': coefficient -1e6 on binary 'b' is far larger in magnitude than the tightest valid value -5
- --> row 'c1'
- = row: c1: x - 1e6 b <= 0
- = help: set the coefficient of b in c1 to -5 and the upper bound to 0 (was -1000000 and 0)
- `rusolve --explain RS2001` explains any code. With `--format json`, the same diagnostics are
- part of the JSON output. The list of codes is in
- [`docs/agent/diagnostics.md`](docs/agent/diagnostics.md).
- Conventions: `--format json` prints exactly one JSON object on stdout (logs go to stderr);
- schemas are in [`docs/schema/`](docs/schema). Exit codes: 0 ran (the solve status is in the
- output), 2 usage error, 3 input error, 4 internal error. A misspelled option or `RUSOLVE_*`
- variable is an error that suggests the nearest valid name.
- ## For AI agents
- - **MCP server** (stdio): `claude mcp add rusolve -- /path/to/rusolve-cli mcp`, or in any MCP
- client `{"mcpServers": {"rusolve": {"command": "/path/to/rusolve-cli", "args": ["mcp"]}}}`.
- Tools: `load_model`, `build_model`, `solve`, `get_solution`, `modify`, `check_solution`,
- `explain`, `lint`, `ranging`, `feasrelax`, `free_model`. Design:
- [`docs/design/mcp.md`](docs/design/mcp.md).
- - **Entry point:** [`llms.txt`](llms.txt). Full generated reference of commands, MCP tools,
- options and environment variables: [`docs/agent/reference.md`](docs/agent/reference.md).
- - **Skills:** task recipes in [`docs/skills/`](docs/skills) (start with `start-here`, then
- `solve`, `debug-infeasible`, `tune`, `extend`, and playbooks for common model classes such as
- scheduling, routing, lot sizing and facility location). The folder is also a Claude Code plugin
- (`.claude-plugin/`).
- - **Evaluations:** [`evals/`](evals) measures how well agents use all of this.
- ## Documentation
- [`docs/README.md`](docs/README.md) is the index.
- ## Contributing
- Contributions are welcome, including AI-assisted ones. See [`CONTRIBUTING.md`](CONTRIBUTING.md).
- Agents start at [`AGENTS.md`](AGENTS.md).
- ## License
- Copyright 2026 CREAIT. rusolve is source-available under the [Business Source License 1.1](LICENSE).
- In short:
- - **Free:** reading, modifying, forking and redistributing the code; research, teaching,
- testing and evaluation; and production use to solve your own (or your company's) optimization
- problems.
- - **Needs a commercial license:** using rusolve to provide optimization to others, as a
- product, a hosted service or consulting. Contact [CREAIT](https://creait.nl).
- - **Becomes open source:** each release converts to the GNU AGPL v3 four years after it is
- published.
- [`LICENSING.md`](LICENSING.md) answers the common questions (can my company use it in
- production? can we sell a product built on it?). The [license text](LICENSE) is what applies.
- rusolve contains code translated from HiGHS (MIT) and other permissively licensed projects; their notices
- are in [`THIRD_PARTY_NOTICES.md`](THIRD_PARTY_NOTICES.md).
Ratings and comments
Sign in to rate this skill.
Sign inFree to use. If it’s helped, pass it on.