claude-flow is a community open-source project (github.com/ruvnet/claude-flow, node, mit) that layers swarm topologies, hooks, and a vector-indexed memory store on top of claude code. it is not published or maintained by anthropic. bernstein is agent-agnostic: 40+ adapters including claude code, codex, gemini cli, aider, ollama, openai agents sdk, cloudflare agents. bernstein's coordinator is deterministic python; claude-flow coordinates inside claude, so planning consumes model tokens. bernstein adds an opt-in hmac-chained audit log, pluggable sandbox backends (worktree, docker, e2b, modal, cloudflare, daytona), an mcp server mode, and an eval harness with a failure taxonomy. pick claude-flow if your stack is claude-code-only and you want its hook and swarm surface. pick bernstein if you want one orchestrator across many cli agents. full matrix at https://bernstein.run/vs/claude-flow.
canonical answer