Get Physics Done (GPD) is an open-source AI copilot designed to automate and structure complex physics research workflows. It assists physicists by transforming research questions into systematic, multi-stage processes, from problem scoping and planning to derivation, verification, and manuscript preparation. GPD is targeted at researchers tackling challenging, long-horizon physics problems that demand rigorous verification, structured memory, and complex analytical or numerical work, aiming to enhance research efficiency and rigor.

How It Works

GPD operates as an agentic AI physicist, guiding research through four distinct stages: Formulate, Plan, Execute, and Verify. It structures work hierarchically into Projects, Milestones, Phases, and Plans, managing dependencies and execution order via "waves." A key feature is its ability to lock research conventions across up to 18 physics fields, ensuring consistency in notation, sign choices, and assumptions throughout a project. GPD integrates with popular AI runtimes, including Claude Code, Gemini CLI, Codex, and OpenCode, providing a unified workflow across different LLM backends.

Quick Start & Requirements

• Primary install command: npx -y get-physics-done
• Prerequisites: Node.js with npm/npx, Python 3.11+ (with venv module), network access to npm and GitHub, and API access for one of the supported AI runtimes (Claude Code, Gemini CLI, Codex, or OpenCode).
• Links: Official quick-start commands and workflow details are available within the README. Runtime-specific help can be accessed via commands like /gpd:help (Claude/Gemini) or $gpd-help (Codex) after installation.

Highlighted Details

• Structured Workflow: Manages research through distinct phases (Formulate, Plan, Execute, Verify) producing concrete artifacts like PROJECT.md, derivation notes, and verification scripts.
• Convention Locking: Enforces consistent notation and assumptions across up to 18 physics fields for multi-stage projects.
• Multi-Runtime Support: Integrates seamlessly with Claude Code, Gemini CLI, Codex, and OpenCode, allowing users to choose their preferred LLM environment.
• Advanced CLI: Offers machine-readable validation, observability, and tracing commands for automation, review, and debugging.

Maintenance & Community

Developed by Physical Superintelligence PBC (PSI), GPD is presented as a community contribution. Specific details on active maintainers, community channels (like Discord/Slack), or a public roadmap are not detailed in the provided README.

Licensing & Compatibility

GPD is released under the Apache License 2.0. This license is permissive and generally compatible with commercial use and linking within closed-source projects.

Limitations & Caveats

The level of detail GPD can expose regarding internal tool and subagent operations is limited by the capabilities of the underlying AI runtime provider. GPD records and emits workflow, session, and trace events it can access locally but does not fabricate opaque provider internals.