Claude Octopus
A Claude Code plugin that runs one task across several external models and surfaces disagreements before you ship
Medium risk
We rate an entry medium when the tool runs code, makes network calls or reads project files. Check what exactly it does before installing.
Why this level
- Runs external models and CLIs, spending tokens and money
- The autonomous mode runs pipeline steps without manual review of each
Install
Manual install
claude plugin marketplace add https://github.com/nyldn/plugins.gitAdds the marketplace, then install the octo plugin via /plugin.
This is third-party code. Review the repository files before installing.
What it does
Claude Octopus adds a multi-model layer to Claude Code: octo commands wire in external providers including Codex, Copilot, Qwen, Ollama, Perplexity, OpenRouter, OpenCode, Cursor CLI, Grok and Kimi Code. Until you call them it stays dormant and does not interfere with ordinary work. When invoked it gathers opinions from different models, runs adversarial review and applies a consensus gate to catch disagreements. It ships a set of role personas, slash commands and reusable skills, with per-run cost tracking.
Who it is for. For developers on Claude Code who want a second and third opinion from other models before important changes.
Good fit when
- You need to cross-check a code or architecture decision across several models
- You want adversarial review and an explicit consensus gate before shipping
- You want to run an autonomous pipeline from spec to result
Not a fit when
- A single Claude Code session is enough with no escalation needed
- You have no access or keys to external providers and want to avoid their cost
Example request
Run this architecture decision across several models and show me where they disagreeLimitations
It needs a recent Claude Code. Multi-model features require access to external providers, some of them paid and key-based; OpenAI and Anthropic APIs need a VPN from Russia. Each multi-model run raises token spend, which is why the plugin includes a cost projection.
How to disable. Uninstall the octo plugin via /plugin and remove the nyldn/plugins marketplace. Without calling octo commands it stays inactive anyway.
Security check
- Runs external models and CLIs, spending tokens and money
- The autonomous mode runs pipeline steps without manual review of each
README in short
The README describes Claude Octopus as an independent project on top of Claude Code for multi-model work. It lists twelve external providers, the consensus gate, adversarial review and an autonomous spec-to-result pipeline. It shows marketplace install for Claude Code, Codex and other hosts, the set of personas, commands and skills, cross-session memory integration and per-run cost tracking. MIT licensed, not affiliated with Anthropic.
FAQ
Do I need all twelve providers?
No. One external provider alongside Claude is enough for the multi-model features. With none you still get personas, workflows and skills.
Does it interfere with normal Claude Code work?
No, not until you call the octo commands. By default it stays inactive and does not intervene in the standard flow.
Related
Open-source personal AI assistant on your own machine: answers in Telegram, Slack, Discord and WhatsApp, extended with skills and plugins
A self-improving agent from Nous Research with a TUI, messaging gateway, cron jobs and skills it writes itself
An open source coding agent for the terminal and desktop with build and plan modes
Open prompt library with a Claude Code plugin, MCP server and CLI: search, fetch and improve prompts and skills from an agent