mexregkan/claude-for-researchers

This repository is a practical guide and toolkit for researchers, particularly physicists and mathematicians, who want to use Claude Code productively on long technical projects. It was written from real experience running a months-long mathematical research project with the tool, covering what actually works, what wastes time, and the failure modes you will encounter. The guide is direct about what Claude Code is and is not. It treats the tool as a capable junior research assistant: it can write, edit, and compile LaTeX quickly, track down obscure errors, write Python or Mathematica computation scripts from a clear specification, navigate large documents without losing structure, and handle tedious mechanical tasks like checking formula cases or renaming things consistently. What it cannot do is notice when a mathematical argument is wrong on its own terms, supply physical intuition, or remember anything between sessions without explicit help. The core of the workflow is the CLAUDE.md file, a document you keep in your project folder that tells Claude everything it needs to know about your work at the start of each session. Without this, every session starts from scratch. The guide explains exactly what to put in it. Alongside that, the repository covers skills (reusable procedures for Claude to follow), a dual-document pattern for LaTeX projects using a full main.tex alongside a condensed version for context management, a next-session-prompts.md file for continuity between sessions, and settings and hooks that automate routine checks. There are also sections on git workflow for academics, how to handle GitHub README pages when your main document is LaTeX, numerical computation patterns, and an honest account of limitations. Templates and scripts included in the repository cover the CLAUDE.md structure and other pieces of the recommended workflow. Installation instructions start from zero: installing VS Code, adding the Claude Code extension, signing in with an Anthropic account, and opening a project folder. The guide notes that a typical research session costs roughly one to five dollars in API tokens depending on how much context is loaded, or is covered by a Claude Pro subscription.