PySR
High-Performance Symbolic Regression in Python and Julia
PySR is an agent skill that enables AI assistants to perform symbolic regression—discovering interpretable mathematical expressions that fit data. Built on a high-performance Julia backend (SymbolicRegression.jl), it searches for equations that optimize objectives while maintaining simplicity and interpretability. Agents can use PySR to convert neural networks into analytic equations, find governing equations from datasets, or distill complex models into human-readable formulas. The skill provides a scikit-learn-style Python interface with support for custom operators, complexity constraints, and export to JAX, PyTorch, or SymPy formats.
Key Features
Use Cases
- 01Discovering physical laws or governing equations from experimental datasets
- 02Converting trained neural networks into interpretable symbolic expressions
- 03Finding compact mathematical models for low-dimensional scientific data
- 04Automated feature engineering by identifying nonlinear relationships
- 05Symbolic distillation for explainable AI in physics-informed machine learning
- 06Equation discovery in domains like astrophysics, dynamical systems, and chemistry
Related Skills
View moresuperpowers
An agentic skills framework & software development methodology that works.
skills
Skills for Real Engineers. Straight from my .agents directory.
skills
Public repository for Agent Skills
ponytail
Makes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote.
PySR — FAQ
What is the PySR agent skill?+
PySR is an agent skill that teaches AI assistants how to perform symbolic regression—finding interpretable mathematical formulas that fit data. It wraps a high-performance Python/Julia library with detailed documentation distilled from forums and docs into a self-contained skill file.
How do I install the PySR skill for my AI agent?+
Download the skill file from the repo and place it in your agent's skills directory. For Claude Desktop or compatible agents using the Agent Skills format, run: mkdir -p ~/.claude/skills/pysr && curl -o ~/.claude/skills/pysr/SKILL.md https://raw.githubusercontent.com/astroautomata/PySR/master/skills/pysr/SKILL.md
Which AI clients work with the PySR skill?+
The skill is designed for AI agents that can read and apply skill files, such as Claude Desktop with the Agent Skills format. Any agent that can execute Python code and follow structured instructions in markdown can use it.
Do I need API keys or prerequisites to use PySR?+
PySR requires Python and will auto-install Julia dependencies on first import. No API keys are needed—it runs locally. You can install via pip (pip install pysr) or conda (conda install -c conda-forge pysr).
Is PySR free to use?+
Yes, PySR is open-source and free. It's available on PyPI, conda-forge, and GitHub under an open license.
What kind of data works best with PySR?+
PySR excels on low-dimensional datasets (typically 5-10 features) where interpretable equations are desired. For high-dimensional problems, techniques like symbolic distillation of neural networks can extend its applicability.
How do I install PySR?+
Open the source repository on GitHub and follow its README. PySR is a skill — MCP Agents Market links you directly to the official repo.
Is PySR free?+
PySR is an open-source project hosted on GitHub. Check the repository for its license and any usage requirements.