home / mcp / area mcp server

Area MCP Server

Provides a consultative MCP workflow to design, validate, preview, and upload battle robots through Firebase-backed endpoints.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "aldegad-mcp-area": {
      "url": "https://<region>-<project-id>.cloudfunctions.net/api",
      "headers": {
        "MCP_ARENA_API_BASE_URL": "https://<region>-<project-id>.cloudfunctions.net/api"
      }
    }
  }
}

You can run Area MCP Server locally or deploy it to a Firebase-backed environment to enable a consultative, scripted approach for creating, validating, previewing, and uploading battle robots within an MCP workflow. It combines a web frontend, Firebase Functions API, and an MCP bridge to connect agents, offering a complete end-to-end cycle from strategy and script design to final upload.

How to use

You interact with the MCP through a client that drives a consultative workflow. The agent helps you define your robot’s strategy, drafts a script, asks follow-up questions to fill in missing details, validates the script, previews a duel, and finally uploads the robot once you approve.

A typical usage sequence is: start a consultative flow, draft the robot’s design, answer follow-up questions to complete required details, validate the script, preview a duel, and, after your confirmation, upload the final script.

How to install

Prerequisites: you need Node.js and Yarn installed on your development machine.

1) Install dependencies for the monorepo.

yarn install

2) Run the Firebase emulators to simulate the hosting, functions, Firestore, and storage locally.

yarn emulators

3) In a separate terminal, start the web UI for uploading robots and running battles.

yarn dev:web

4) Open the web app in your browser to upload robots, run battles, and view replays.

http://localhost:3000

Additional notes and configuration

Agent connections use a stdio MCP bridge. Start it in a separate terminal so the agent can connect to the local MCP API.

5) To connect an agent via the MCP bridge, run the bridge command.

yarn dev:mcp-bridge

Deployment considerations

When you are ready to deploy to a real Firebase project, switch the API base URL to your cloud function URL and deploy hosting, functions, Firestore, and storage.

A typical deployment sequence is to build the web assets and then deploy all Firebase components.

yarn build:web
firebase deploy --only hosting,functions,firestore,storage

Available tools

get_build_flow

Guides the agent on the sequence of consultative steps to determine how to build the robot workflow.

coach_robot_design

Generates a draft robot design based on user intent and preferences, returning rules and initial scripts.

validate_robot_script

Checks the robot DSL for syntax and tactical quality, returning warnings or recommendations.

preview_robot_duel

Runs a preview duel to evaluate candidate scripts and show expected outcomes.

upload_robot_script

Handles the final upload of the robot script after user approval.