consensus
Break through difficult problems by gathering diverse AI perspectives in parallel. Consensus queries multiple AI providers concurrently — GP
Open source Repository Open in the app JSON README (API)
About
Break through difficult problems by gathering diverse AI perspectives in parallel. Consensus queries multiple AI providers concurrently — GPT-5.2, Gemini 3.1 Pro, and DeepSeek V3.2 by default — then synthesizes their responses to surface agreements, unique insights, and actionable takeaways. Choose from any model available on OpenRouter, or route providers directly through their native APIs (OpenAI, Gemini). Add, remove, and configure providers interactively with natural language.
Details
- Kind
- Plugins
- Topic
- AI, RAG & memory
- Publisher
- mark-kimura
- Origin
- marketplace
- Category
- ferramentas
- Last push
- 2026-05-07T16:12:59Z
- Repository state
- ativo
- Language
- Python
- License
- MIT
- Added
- 2026-08-30 01:48:58
- Updated
- 2026-08-30 01:48:58
- Origin id
mark-kimura/consensus-plugin/consensus
README
# Consensus — Multi-AI Provider Plugin for Claude Code Get unstuck by querying multiple AI providers (GPT-5.5, Gemini, DeepSeek V4 Pro) for diverse perspectives, right from within Claude Code. ## How It Works When you're stuck on a problem, Consensus queries multiple AI providers concurrently — GPT-5.5, Gemini 3.1 Pro, and DeepSeek V4 Pro via OpenRouter by default — then synthesizes their responses to surface common themes, unique insights, and actionable takeaways. ## Installation ### 1. Install `uv` (Python package runner) ```bash curl -LsSf https://astral.sh/uv/install.sh | sh ``` ### 2. Set API Keys Only one key is needed for the default setup: ```bash export OPENROUTER_API_KEY="sk-or-..." # Routes all three default providers ``` Optional, for direct API access (bypassing OpenRouter): ```bash export OPENAI_API_KEY="sk-..." # OpenAI direct API export GEMINI_API_KEY="AI..." # Google Gemini direct API ``` Add them to your shell profile (`~/.bashrc`, `~/.zshrc`, etc.) to persist across sessions. ### 3. Install the Plugin **Option A — Third-Party Marketplace**: ``` /plugin marketplace add mark-kimura/kahidreamers-marketplace /plugin install consensus@kahidreamers-marketplace ``` **Option B — Official Plugin Directory** (not yet available — submitted, under review by Anthropic): ``` /plugin install consensus@claude-plugin-directory ``` ### 4. Enable Auto-Update (Recommended) For the official directory, auto-update is enabled by default. For third-party marketplaces, enable it manually: 1. Run `/plugin` in Claude Code 2. Go to **Marketplaces** 3. Select **kahidreamers-marketplace** 4. Choose **Enable auto-update** Without this, you'll need to run `/plugin update consensus@kahidreamers-marketplace` manually after each release. ## Usage ### `/consensus:ask` ``` /consensus:ask help me optimize the SSE connection handling ``` Or without arguments — Claude infers from conversation context: ``` /consensus:ask ``` This also auto-triggers when you say things like: - "What do other AIs think about this approach?" - "Can I get a consensus on this?" - "Cross-check this with other models" ### `/consensus:config` ``` /consensus:config # Show current providers (same as check) /consensus:config update # Update all providers to latest models /consensus:config add llama 4 # Add a new provider via OpenRouter /consensus:config remove deepseek # Remove a provider /consensus:config openai # Interactively configure OpenAI (model + routing) /consensus:config --reset # Reset all providers to plugin defaults ``` ### What Happens 1. Claude analyzes your request and conversation context 2. Crafts a comprehensive prompt with technical details 3. Queries all available providers concurrently 4. Synthesizes findings: agreements, unique perspectives, and takeaways All intermediate files go to a temp directory — nothing is written to your project. ## Configuration The plugin works out of the box with just `OPENROUTER_API_KEY`. Use `/consensus:config` to manage providers interactively, or customize `consensus_config.json` directly: - **Provider routing**: Route through OpenRouter or call APIs directly (`/consensus:config openai`) - **Models**: Add, remove, or update providers (`/consensus:config add llama 4`) - **Search modes**: Set default to `"web"` (includes web search) or `"none"` (faster, no web) - **Token limits**: Adjust `max_tokens` per provider ### Config Resolution Order 1. `./consensus_config.json` (project root — for project-specific overrides) 2. `~/.claude/consensus_config.json` (user config — written by `/consensus:config`, survives plugin updates) 3. Plugin's built-in `consensus_config.json` 4. Hardcoded defaults ## Search Modes By default, providers include current web information in their responses. If you'd rather get purely model-based perspectives without web search, just say so naturally: - "Get a consensus on this, no web search" - "What do other models think — don't use the web" - "/consensus:ask without web search, compare approaches to caching" Claude will detect your intent and run in the appropriate mode: - **`web`** (default): Providers include current web information. Good for questions about recent technologies or current best practices. - **`none`**: Pure model knowledge, no web search. Faster responses. Good for coding questions, architecture decisions, or when you want unbiased model perspectives. ## Project Structure ``` consensus-plugin/ ├── .claude-plugin/ │ └── plugin.json # Plugin manifest ├── commands/ │ └── config.md # /consensus:config command ├── skills/ │ └── ask/ │ └── SKILL.md # /consensus:ask skill (also auto-triggers) ├── hooks/ │ └── hooks.json # SessionStart: checks API keys + uv ├── scripts/ │ ├── consensus.py # Core engine (uv run auto-resolves deps) │ └── check-setup.sh # Setup validator ├── consensus_config.json # Default config template ├── .gitignore ├── LICENSE └── README.md ``` ## Requirements - [uv](https://docs.astral.sh/uv/) — auto-installs Python dependencies on first run - At least one API key (`OPENROUTER_API_KEY` recommended) - Claude Code ## License MIT