zts212653/clowder-ai

Clowder AI is a platform layer that orchestrates multiple AI agents into a coordinated team rather than isolated tools.

The problem it addresses is the friction of working with multiple AI models simultaneously. When developers use Claude, GPT, Gemini, and other models together, they become the routing layer themselves, manually copying context between tools, tracking conversation state, and managing handoffs. Clowder AI solves this by providing persistent agent identity, cross-model review workflows, shared memory across sessions, and agent-to-agent communication with structured handoffs. Each agent maintains its role and personality across interactions, and the platform handles the coordination that would otherwise require manual middle management.

Clowder AI suits teams that want to leverage multiple models' strengths on the same problem—having Claude handle architecture while GPT reviews code, for example. It works with existing agent CLIs like Claude Code, Codex CLI, Gemini CLI, and opencode rather than replacing them, sitting as a coordination layer above those tools. The platform includes an evidence store for shared knowledge, async messaging with mention-based routing, and a plugin framework supporting MCP tools and instant messaging adapters. Desktop installers for Windows and macOS are available, and the tool can run from source with Node.js 20 and pnpm 9, with Redis as an optional dependency.

The project originated from Cat Cafe, a production workspace where multiple AI agents collaborate on real software, meaning every feature has been tested in actual use. Development appears active with documentation spanning setup guides, architecture decisions, and feature specifications across multiple languages.