The AI Counsel MCP
A local web service and MCP server that gathers several models to answer, peer review each other or debate as advisor personas.
- 142 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Questions and keys go to several external providers and costs follow your plans.
- Keys are kept in a local file as plaintext.
Install
- Works in
Manual install
git clone https://github.com/jacob-bd/the-ai-counsel.git && cd the-ai-counsel && uv sync && npm install --prefix frontend && ./start.shFirst command from the Installation section of the README.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Questions and keys go to several external providers and costs follow your plans.
- Keys are kept in a local file as plaintext.
This is third-party code. Review the repository files before installing.
What it does
The AI Counsel runs in two modes. In Council mode several models answer a question independently, anonymously peer review each other's answers and a chairman model synthesizes the result. In Advisors mode named personas such as the Skeptic, Strategist and Ethicist debate over several rounds and deliver a structured verdict with an action plan. The backend is FastAPI with a React interface, with twelve providers, web search and PDF and text file uploads. The MCP server gives Claude Code and Gemini CLI access to an instance, including a `run_iterative_debate` tool. It runs locally or in Docker, and a `the-ai-counsel-api` skill describes a REST fallback.
Who it is for. People making contested decisions who want to compare several models' answers. Needs Python 3.10 or newer, Node.js 18 or newer and provider keys.
Good fit when
- A question has trade-offs and needs a debate between roles.
- One model's answer is checked by several others' review.
- An agent in Claude Code runs a debate through MCP and collects the result.
- PDFs and text files are attached to the discussion.
- The service is deployed on your own server in Docker.
Not a fit when
- A simple factual answer is needed, since several models cost more and run slower.
- The question cannot go to third-party providers.
- The instance would be exposed to the internet without protection.
Example request
Run an advisor debate on whether to rewrite the module in Rust and return the verdict with a plan.MCP
- Transport
- stdio, sse
- Authentication
- API key
Limitations
Each request goes to several external providers and is billed through your keys, while multi-round debates multiply cost. Keys live in `data/credentials.json` as plaintext with restricted permissions or in the OS keystore on desktop, with no OS keystore in Docker. `LLM_COUNCIL_BIND_HOST=0.0.0.0` exposes the service on the network, so shield it from outsiders. The synthesized answer is still model output and needs checking.
How to disable. Remove the MCP entry with `claude mcp remove the-ai-counsel`, stop the servers or the container, and delete the `data` folder with keys and history.
FAQ
How does Advisors differ from Council?
Council gives one synthesized answer after peer review. Advisors runs a persona debate in rounds and ends with a verdict and plan. Pick the second for contested decisions.
How is MCP connected?
Install the package with `pip install -e .` and run `claude mcp add the-ai-counsel python -m the_ai_counsel_mcp`. A remote SSE option exists.
Where are keys stored?
In the local `data/credentials.json` or the OS keystore on desktop. Docker has no OS keystore.
How is it run in Docker?
Run `docker run -d --restart unless-stopped -p 8001:8001 -v ./data:/app/data ghcr.io/jacob-bd/the-ai-counsel:latest`.
What is needed to run it?
Python 3.10 or newer, Node.js 18 or newer and the `uv` manager. Provider keys go into Settings after launch at `http://localhost:5173`.
Discussion
No questions or reviews yet.
Related
MCP serversbrainstorm-mcp runs debates among GPT, Gemini, DeepSeek, Claude and Ollama models in your editor and returns a three-point verdict
A Claude Code plugin that asks several AI models the same question and shows their answers side by side with a synthesis of agreement and divergence
An MCP server that lets Claude Code, Codex and other agents call Gemini, GPT, Grok and local models for review, debugging and consensus
A plugin that gives an agent the IntelliNode skill and MCP server: ask any model, search, review code, tests, SQL, images and speech
Plugin and MCP server that asks GPT, Gemini, Grok and local models to review a plan or hunt for bugs before code ships.
An MCP server that sends one prompt to several OpenAI, Anthropic, Gemini, Groq, DeepSeek and Ollama models at once and can draw a decision from them
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| The AI Counsel (this card) | no | API key | 142 | |
| LLM Council | no | 3,015 | three stages of answers, anonymous review and chairman; here also a persona debate mode with voting | |
| claude-council | no | 932 | a council of models inside Claude Code; here a separate web service with an interface, Docker and MCP |
README in short
The interface, MCP and REST expose one service, so debates can run from a browser or an agent. Documentation is kept in sync between them.