Agent Memory Compiler
Write your project's rules once — get AGENTS.md, CLAUDE.md, Cursor, Copilot and Gemini memory files, instantly. Free, no signup, and it warns you when a file gets too long (long context files hurt agents).
Fill in the form (or “Load an example”) to compile your agent memory.
Keep it in sync — don't hand-maintain five files
The catch with generated files is drift: edit one, forget the rest. In TaskPrio your project's rules live in one place and your agents pull them over MCP with get_next_task — so the memory and the work queue are the same source of truth. Edit once; every tool and every agent stays current.
How it works
One project, one set of rules, five tool-native files. AGENTS.md is the canonical output (read by 28+ agents); CLAUDE.md, .cursor/rules/*.mdc, .github/copilot-instructions.md and GEMINI.md are the same guidance placed where each tool looks. Drop each file at its path and your agents read it automatically.
Keep it lean (this matters)
Bigger isn't better. ETH Zurich's Evaluating AGENTS.md found machine-generated, dump-everything context files lowered task success ~3% and raised cost 20%+, while short hand-curated ones improved success ~4%. So this tool flags any file over a ~1500-token budget — keep your highest-priority rules and cut the rest.
FAQ
What is AGENTS.md?
An open, tool-agnostic Markdown file telling AI coding agents how to work in your repo — stack, commands, conventions, what not to do, definition of done. Read by 28+ tools, so it's the canonical output here.
How is CLAUDE.md different?
Same guidance, different location — CLAUDE.md is where Claude Code looks. Cleanest is one source: keep AGENTS.md and ln -s AGENTS.md CLAUDE.md. This tool emits a ready CLAUDE.md (with that tip) either way. Full breakdown: AGENTS.md vs CLAUDE.md →
Which tools read which file?
AGENTS.md → Codex / Aider / Cline / Continue / Goose / 28+. CLAUDE.md → Claude Code. .cursor/rules/*.mdc → Cursor. .github/copilot-instructions.md → GitHub Copilot. GEMINI.md → Gemini CLI.