claude-octopus

Why we included this project

For teams that already live inside Claude Code, this plugin sends a single task to several external models at once (Codex, Gemini, Copilot, Qwen, Ollama, and more) and compares their output against Claude's. A 75% consensus gate flags the disagreements, so a confident but wrong answer is more likely to get caught before it ships. That makes it a good fit for security reviews and other high-risk work where a second opinion is worth more than the setup cost. The plugin routes each request to the right workflow and, with claude-mem or agentmemory, carries decisions and context across sessions, so recurring reviews don't start from scratch. If you trust one model most days but want extra voices on the risky calls, this gives you that escalation without leaving Claude Code.