Free tool · no signup · runs in your browser

CLAUDE.md generator

A good context file is the highest-leverage five minutes in agentic coding: commands the agent can't guess, conventions it would otherwise violate, lines it must not cross. Fill in what's true for your repo — the file assembles live. Not sure about the format? CLAUDE.md vs AGENTS.md, explained.

1 · Project

2 · Commands

3 · Conventions

4 · Guardrails

5 · Format

CLAUDE.md
  

Context file FAQ

What should a CLAUDE.md file contain?

The things an agent cannot guess and will otherwise get wrong: the exact build, test, and lint commands; project conventions that differ from defaults; paths it must not touch; and etiquette rules like "ask before adding dependencies." Keep it short — a screenful the agent actually follows beats three pages it skims.

CLAUDE.md or AGENTS.md — which should I generate?

AGENTS.md is the cross-tool convention read by many agent CLIs; CLAUDE.md is Claude Code’s native file. If you only run Claude Code, CLAUDE.md alone is fine. If you run several agents, put the content in AGENTS.md and generate the one-line CLAUDE.md pointer this tool offers — one source of truth, every tool reads it.

Where does the file go?

The repo root is read automatically at session start. Subdirectories can carry their own CLAUDE.md for area-specific rules — agents read the closest one. Commit it like any other file so the whole team (and every agent) shares the same context.

Does this tool send my project details anywhere?

No. The generator runs entirely in your browser — there is no backend, no analytics event on your answers, and nothing to sign up for. Copy the output and paste it into your repo.

A context file tells your agent how to work; the other half is noticing when it needs you. Backgrind runs your CLI in an always-on-top overlay and pings you only on decisions — and the notification builder wires up the hooks in a minute.