How do you move Cursor rules to Claude Code, Copilot or Codex?

Short answer

Map by rule type. Always Apply rules go to AGENTS.md or CLAUDE.md, glob rules to .claude/rules/ with a paths field (or .github/instructions with applyTo for Copilot), Apply Intelligently rules become skills, and Apply Manually rules become skills you call with /name. Cursor also reads .claude/skills and CLAUDE.md, so one set can serve both tools.

Map by rule type, not by file

A .cursor/rules folder holds four kinds of rules that only look alike. The trick in moving them is to sort them by mode first, because each mode has a different home in the other tools:

Cursor rule Claude Code GitHub Copilot Codex
Always Apply CLAUDE.md, or a file in .claude/rules/ without paths .github/copilot-instructions.md AGENTS.md
Apply to Specific Files (globs) .claude/rules/name.md with paths: .github/instructions/name.instructions.md with applyTo: A nested AGENTS.md in that folder
Apply Intelligently (description) A skill in .claude/skills/name/SKILL.md An Agent Skill with a description An Agent Skill with a description
Apply Manually (@-mention) A skill with disable-model-invocation: true, run as /name Invoke an installed skill explicitly; controls depend on the client Invoke an installed skill explicitly with $name
User Rules ~/.claude/CLAUDE.md or ~/.claude/rules/ Not covered here AGENTS.md in ~/.codex
AGENTS.md Read directly, or imported with @AGENTS.md Read by Copilot’s agents Read natively

Two notes on the Copilot column. Path-specific instruction files with applyTo are, on GitHub.com, only supported for Copilot cloud agent and Copilot code review. And for Codex a nested AGENTS.md is a folder-level approximation of globs, not a file-type filter: Codex reads the files from the project root down to the directory it works in, and later files override earlier ones.

The move, step by step

  1. Sort your rules. Open each .mdc file and note its mode: alwaysApply: true, globs, only a description, or none of these.
  2. Shared always-on rules to AGENTS.md. This is the one file Cursor, Codex and Copilot’s agents read, and Claude Code reads it too. If the repository also has a CLAUDE.md, Claude Code reads only the CLAUDE.md by default, so add the line @AGENTS.md to it.
  3. Glob rules to .claude/rules/. Rewrite the frontmatter: globs: src/api/**/*.ts becomes a paths: list. A path-scoped rule loads when Claude reads or edits a matching file. For Copilot, make the same rule a .instructions.md file with applyTo.
  4. Description rules to skills. Make a folder .claude/skills/<name>/ with a SKILL.md. The description field does the same job as in Cursor: it is what the agent reads to decide whether to load the skill. The body only loads when used, so long procedures cost little.
  5. Manual rules to commands. Same as step 4, plus disable-model-invocation: true, so the skill runs only when someone types /name.
  6. User Rules by hand. They are stored on your Cursor account, not as files. Copy the text into ~/.claude/CLAUDE.md, and for Codex into ~/.codex/AGENTS.md.
  7. Test. Run /memory in Claude Code to list and open your instruction files, then ask Claude to read a file that should trigger a path rule and check that the rule is followed.

Cursor’s /migrate-to-skills command can help with step 4: it turns Apply Intelligently rules into skills in .cursor/skills/. It skips always-on rules, glob rules and User Rules. Move the output to .claude/skills/ so both tools find it.

The common mistake: copying the frontmatter along

The most common failure is copying .mdc files into .claude/rules/ and renaming them. Claude Code reads exactly one field from a rule, paths; every other field, including globs, alwaysApply and description, is ignored without an error. A rule without paths loads in every session. So ten carefully scoped Cursor rules silently become ten always-on rules, and the context fills up. Rewrite globs as paths, and turn description rules into skills instead of rules.

The reverse surprise is pleasant: you may not need two copies. Cursor reads CLAUDE.md and always applies it, and it loads skills from .claude/skills/ for compatibility. Skills in both tools also accept a paths field to limit when they are surfaced. A repository with AGENTS.md, a short CLAUDE.md that imports it, and skills in .claude/skills/ serves Cursor and Claude Code from one set of files. Only glob-scoped rules still need a form per tool.

For the plain question of what belongs in AGENTS.md versus Cursor rules, see Cursor rules vs skills and What is AGENTS.md.

What this means for a team

Most teams do not switch tools at once; one developer tries Claude Code while the rest stay in Cursor, someone else uses Copilot. Every rule that exists in only one tool’s format is a rule the others do not get. Moving to the shared formats, AGENTS.md and Agent Skills, is less about migration and more about making the team’s agreements tool-independent. What does not move are the things neither format holds: your Cursor chat history, and anything Agent learned that was never written down.

Where Bombos fits

Converting rules between formats keeps written instructions in step, but each tool still keeps its own context. Bombos is a plugin that gives every AI tool your team uses, Cursor, Claude Code, Copilot, Codex and your own agents, one shared memory with team knowledge and rules learned from corrections, traceable, in the cloud or on your own server. One developer can try it first while switching tools, so the context comes along.

Related questions

Can I just rename my .mdc files to .md and drop them in .claude/rules?

They will load, but not as you expect. Claude Code reads only the paths field from a rule and ignores other fields, including globs and alwaysApply. A rule without paths loads in every session, so your scoped rules become always-on until you rewrite globs as paths.

Does Claude Code read AGENTS.md?

Yes, from v2.1.277, but by default only when there is no CLAUDE.md or CLAUDE.local.md in your working directory or above it. If you have both, put @AGENTS.md in your CLAUDE.md.

Do Claude skills work in Cursor?

Cursor loads skills from .claude/skills/ and ~/.claude/skills/ for compatibility, next to its own .cursor/skills/ and .agents/skills/. Both tools follow the Agent Skills format with a SKILL.md and a description.

Where do my Cursor User Rules go?

In Claude Code, personal instructions for all projects go in ~/.claude/CLAUDE.md or ~/.claude/rules/. In Codex, the global file is AGENTS.md in ~/.codex. Cursor User Rules are not files, so copy them by hand.

Sources

  1. Cursor docs: Rules
  2. Cursor docs: Agent Skills
  3. Claude Code docs: How Claude remembers your project
  4. Claude Code docs: Skills
  5. GitHub Docs: Adding repository custom instructions for Copilot
  6. OpenAI Codex docs: AGENTS.md
  7. GitHub Docs: Adding agent skills for Copilot
  8. OpenAI Codex docs: Skills and plugins
The first half hour · free

What does your AI forget?

  1. 01You tell us which AI you use and where it keeps starting over.
  2. 02We show what a shared memory changes there, on your own examples.
  3. 03You keep the plan, with or without Bombos.
Piet Baudoin
Piet BaudoinYou talk to Piet, from the Bombos teamDutch team from Utrecht and Nijmegen · data in Europe
We reply the same working day.