A Model Context Protocol server that connects Squad (the AI-powered product-discovery and strategy platform) to any MCP-aware large-language-model (LLM) application.

It exposes a rich tool-kit for creating, querying and updating product-strategy artefacts directly from your favourite AI co-pilot, so you can research, ideate and plan products in one conversational flow, without ever leaving your editor or chat window.

👉 Download Squad MCP 👈

Why this matters for product teams

Bottom line: MCP turns Claude into an extension of Squad, giving PMs instant, conversational access to strategy artefacts while developers stay in sync with up-to-date requirements.

Under the hood

Our open-source squad-mcp server bundles a curated set of tools generated from the Squad API:

All tools conform to the MCP JSON schema format so that agents can introspect inputs and outputs automatically.

🚀 Quick start

1 · Obtain a Squad API key

1. Sign up / sign in at https://meetsquad.ai.

2. Open Settings → API Keys.

3. Create a key and copy the value.

   

2 · Run the server/download mcp

Choose whichever installation method suits your environment.

Option A – Stand-alone executable

Download Squad MCP from the latest binary for your OS from the project’s GitHub releases page.

# macOS / Linux — make the file executable first
chmod +x squad-mcp
SQUAD_API_KEY=<your-key> ./squad-mcp

# Windows

Option B – Docker

# Clone the repo
git clone https://github.com/the-basilisk-ai/squad-mcp.git

# Move into the repo directory
cd squad-mcp

# Build the image (once)
docker build -t mcp/meet-squad -f Dockerfile .

# Run the server on stdio
docker run --rm -i \
  -e SQUAD_API_KEY

Option C – From source

git clone https://github.com/the-basilisk-ai/squad-mcp.git
cd squad-mcp
npm install
npm run build            # transpiles to ./dist
node

⚙️ Integrating with an MCP client

Add a mcpServers entry to your client’s configuration (claude_desktop_config.json, Cursor, etc.) and restart the client.

Using the stand-alone executable

{
  "mcpServers": {
    "meet-squad": {
      "command": "C:/path/to/squad-mcp.exe",
      "env": {
        "SQUAD_API_KEY": "YOUR_API_KEY_HERE"
      }
    }
  }
}

Using Docker

{
  "mcpServers": {
    "meet-squad": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "-e", "SQUAD_API_KEY",
        "mcp/meet-squad"
      ],
      "env": {
        "SQUAD_API_KEY": "YOUR_API_KEY_HERE"
      }
    }
  }
}

(Prefer "command": "npx" if you installed via NPX.)

Once your client restarts you’ll see the Squad tools (hammer 🔨 icon) listed and ready for use.

🛠️ Environment variables

Example workflow: from idea to production

1. Ideate in Claude – “Create an opportunity for ‘users abandoning checkout at the payment step’.”

2. Claude runs opportunity_create, then suggests three solutions.

3. You choose one; Claude calls solution_create and requirement_create.

4. You see a solution appear in your Squad workspace and requirements in the backlog – ready for engineering.

You’ve gone from a vague hunch to developer-ready solutions without opening a browser.

FAQ

How secure is this?

The server only talks to https://api.meetsquad.ai with your personal API key and streams responses back to the client. No data is persisted outside Squad.

Can I self-host?

Yes – run the stand-alone binary or Docker container on any machine your client can reach.

Does it work with other MCP clients?

Absolutely. Any client that supports MCP 0.2 + can use the same server.

Where can I contribute or file issues?

GitHub: https://github.com/the-basilisk-ai/squad-mcp. PRs welcome! 🎉

The evolving role of PMs in an AI-driven future

Modern PMs spend too much time shuttling information between tools instead of shaping product vision. By wiring Squad into Claude (or any LLM agent) via MCP, your strategic context becomes programmable – you can interrogate it, mutate it and watch the downstream artefacts update in real time. That frees you to focus on why we build, not on copy-paste busywork.

Give it a spin and tell us what you automate first!