Ruflo: a harness around Claude Code and Codex

Ruflo calls itself an agent meta-harness for Claude Code and Codex. The idea is simple even if the repo is big.

ruvnet/ruflo is MIT licensed with about 77k stars. Its README frames it as Agent = Model + Harness: the model writes, and the harness supplies tools, memory, loops and controls. Ruflo is the harness around Claude Code or Codex. It adds, per the README, 100+ specialized agents, swarms that coordinate, and memory across sessions.

In Claude Code you install it as plugins: /plugin marketplace add ruvnet/ruflo, then /plugin install ruflo-core@ruflo. It needs Claude Code 2.1.287 or later. With other MCP clients you run npx -y ruflo@latest mcp start.

Two cautions. The README says its mods run with your account's permissions, so review their code before installing. And the repository is about 650 MB, so a casual clone is a commitment.

I like that it ships an optional terminal view, npx ruflo tui, to show what the agents are doing. With any harness like this, the test is whether you can still see what it's doing.

Would you trade a simpler setup for 100 specialized agents?