To share context between the coding agents on a team, split it in two. Instructions (how to build, conventions, rules) go in a file committed to the repository: AGENTS.md is read by Codex, Cursor and
To share Cursor rules with a team, commit them to the repository as project rules in .cursor/rules/ (or a plain AGENTS.md), and use Team Rules from the Cursor dashboard for rules that apply to every r
Put instructions in CLAUDE.md and history in shared memory. CLAUDE.md (or AGENTS.md) is loaded in full into every session, so it should hold short, stable rules: how to build, conventions, where thing
Claude Code doesn't remember your last session: each one starts with a fresh context window. Three things carry knowledge over. CLAUDE.md holds the instructions you write. Auto memory holds notes Clau
AI coding agents repeat reverted changes because the reason for the revert isn't anywhere they look. The code shows only what is there now, not what was tried and taken out. To stop it, write the reas