Meta Council Now Has an MCP Server: Any AI Agent Can Convene Expert Panels

Meta Council Now Has an MCP Server: Any AI Agent Can Convene Expert Panels

The new customer for infrastructure is not a human with a browser. It is an LLM with a tool-call interface.

We have been thinking about this shift for months. When Daniel Miessler said at [un]prompted that "your company exists as an API, and if people's AIs can't use your company in that way, you kind of don't exist" -- that was the trigger to build what we shipped today.

What We Built

Meta Council now has an MCP server that exposes the entire platform as tools for AI agents. Any MCP-compatible client -- Claude Code, Claude Desktop, Cursor, Windsurf, Cline -- can now:

1. Convene an expert council on any question, with 280+ specialized AI agents 2. Choose from 23+ pre-built panels spanning biotech, finance, software engineering, crisis response, military strategy, and more 3. Configure which LLM model the agents use (Claude, GPT, Gemini, open-source) 4. Manage API keys for premium tools (stock data, AI search) and LLM providers 5. Get the full synthesis -- executive summary, risk matrix, dissenting views, and action plan

The setup is one command:

pip install meta-council-mcp
export META_COUNCIL_API_KEY="mc_your_key"
claude mcp add meta-council -- meta-council-mcp

Then you just talk to your AI assistant naturally: "Convene the biotech panel to analyze whether we should proceed with our Phase IIb trial given the marginal endpoint miss."

Your AI agent calls convene_council, waits for 5-7 experts to deliberate independently, and returns a structured decision document with confidence scores and explicit disagreements.

Why This Matters

Most AI tools are designed for humans using browsers. You click buttons, fill forms, read dashboards. But the fastest-growing consumer of software is not a person -- it is another AI agent executing a task.

When an AI coding assistant needs a second opinion on an architecture decision, it should not tell the user to "go check Meta Council." It should call Meta Council itself, get the expert panel's analysis, and incorporate it into its recommendation. That is what MCP enables.

The Tool-Calling Stack

We did not just ship an API wrapper. The agents themselves now have tools:

- A Financial Controller agent can call a calculator, look up SEC filings, check exchange rates, and pull stock data -- all during its analysis - A Biostatistician can search PubMed for relevant clinical trials before forming an opinion - A Software Architect can search arXiv for the latest research papers on the architecture pattern you are evaluating

There are 11 tools available today. 10 are completely free (web search via DuckDuckGo, Wikipedia, SEC EDGAR, PubMed, arXiv, exchange rates, weather, calculator, date/time, URL fetch). Premium tools use a bring-your-own-key model -- you add your Alpha Vantage key for stock data, your Tavily key for AI search, and the agents use them automatically.

Dynamic Panels: AI-Curated Experts

For Pro users, we also shipped something we have not seen anywhere else: dynamic panel generation. Instead of choosing from pre-built panels, you can let the system analyze your query and select the 5 most relevant, non-redundant experts from our pool of 280+ agents.

It uses Maximum Marginal Relevance (MMR) over semantic embeddings of every agent's expertise profile. Each query gets a unique panel -- a combination of experts that has never been assembled before and is specifically tailored to the dimensions of your question.

The Bigger Picture

We are building for a world where AI agents are the primary consumers of decision infrastructure. The MCP server is our first step toward that future. But the implications go further:

- Composability: An AI agent running a complex workflow can convene a Meta Council at any decision point, not just when a human asks - Specialization: Instead of one general-purpose model trying to be an expert in everything, Meta Council provides access to 280+ specialized perspectives - Grounded analysis: Tool-calling agents do not just opine -- they look things up, verify facts, and run calculations before forming recommendations

The builders who read the stack early build the companies that define the next era. We think multi-expert decision intelligence, accessible via tool calls, is a layer worth building on.

---

Try it at meta-council.com. See the API docs to integrate the MCP server with your tools.