FAQ

FAQ

Is anycrew a framework like CrewAI or LangGraph? No. Those are runtimes for building agent systems inside your product. anycrew is a crew of agents that help you build your product inside the coding assistant you already use. The ai-engineer agent and /crew-llm help you design a product that uses such a runtime.

Do I have to use the whole method? No. Each skill works alone: /crew-review on a diff, /crew-security on a repo, /crew-adr for one decision. The gates and IDs pay off most when the whole loop is used.

Will it fight my existing CLAUDE.md, rules or agents? init never overwrites AGENTS.md, anything under docs/, or any file without the anycrew:generated marker. If a CLAUDE.md exists it is left alone and reported; add @AGENTS.md to it by hand. Your own agents coexist with the crew.

Why Markdown and JSON, no YAML config? So that uvx anycrew and npx anycrew run with zero dependencies on any machine and in CI. See ADR-003.

Why are skills prefixed crew-? They land in .agents/skills next to your own and other packs'. review or plan would collide.

Which assistant works best? Claude Code and Antigravity are verified end to end. Codex, Cursor, OpenCode and Copilot outputs follow their documentation and are being exercised next.

Can I use a cheaper model everywhere? Yes: override the tier mapping in .anycrew/config.json under models. The tiers only say what the role needs; the mapping says what you are willing to pay.

How do I update the crew in a project? Upgrade the package and run anycrew sync. Generated files are rewritten; your edits to unmarked files are kept and reported.

Does it send any data anywhere? No. anycrew is files and a local CLI. The assistants you run it in have their own policies.