ResQ MCP: Disaster Response Intelligence for AI

A production-ready Model Context Protocol (MCP) server that connects AI agents to the ResQ platform's robotics, physics simulations, and disaster telemetry.

Key Features

• Drone Fleet Command — Real-time telemetry, sector scanning, and autonomous swarm deployment via the Hybrid Coordination Engine (HCE).
• Predictive Intelligence — Probabilistic disaster forecasting and sector-level vulnerability mapping (PDIE).
• Digital Twin Simulations — Physics-based RL optimization strategies for incident response (DTSOP).
• Safe-Mode Execution — Built-in protection prevents destructive platform mutations by default.

Requirements

• Python 3.11 or newer
• VS Code, Cursor, Claude Desktop, or any other MCP client

Getting Started

Install the ResQ MCP server with your client. The standard config works in most tools:

{
  "mcpServers": {
    "resq": {
      "command": "uvx",
      "args": ["resq-mcp"],
      "env": {
        "RESQ_SAFE_MODE": "true"
      }
    }
  }
}

{
  "servers": {
    "resq-mcp": {
      "type": "stdio",
      "command": "uvx",
      "args": ["resq-mcp"],
      "env": {
        "RESQ_SAFE_MODE": "true"
      }
    }
  }
}

{
  "mcpServers": {
    "resq": {
      "command": "uvx",
      "args": ["resq-mcp"],
      "env": {
        "RESQ_API_KEY": "your-prod-token",
        "RESQ_SAFE_MODE": "true"
      }
    }
  }
}

claude mcp add resq -- uvx resq-mcp

{
  "mcpServers": {
    "resq": {
      "command": "uvx",
      "args": ["resq-mcp"]
    }
  }
}

{
  "mcpServers": {
    "resq": {
      "command": "docker",
      "args": [
        "run", "-i", "--rm", "--init",
        "-e", "RESQ_SAFE_MODE=true",
        "resqsoftware/mcp"
      ]
    }
  }
}

docker pull ghcr.io/resq-software/pypi:latest

docker run -d --rm --init \
  --name resq-mcp \
  -p 8000:8000 \
  -e RESQ_SAFE_MODE=true \
  resqsoftware/mcp

{
  "mcpServers": {
    "resq": {
      "url": "http://localhost:8000/mcp"
    }
  }
}

docker build -t resq-mcp .

git clone https://github.com/resq-software/pypi.git
cd mcp
uv sync
uv run resq-mcp

Configuration

Control server behavior via environment variables or a .env file:

Note: RESQ_API_KEY defaults to resq-dev-token for local development. No external token is needed to run the server — it works out of the box.

Security & Safety

Safe Mode is enabled by default (RESQ_SAFE_MODE=true). In this state, any tool that performs platform mutations (e.g., dispatching a drone swarm or starting a high-fidelity simulation) will raise a FastMCPError. This allows AI agents to plan missions safely without triggering real-world consequences. Disable this only when you are ready for autonomous execution.

Tools

Mission Control (HCE)

• validate_incident — Submit a confirmation or rejection for an incident report. Supports idempotent re-submission and conflict detection for opposing validations.

  • Parameters: incident_id, is_confirmed, validation_source, correlated_pre_alert_id (optional), notes

• update_mission_params — Push authorized mission parameters to a specific drone for an approved strategy. Includes urgency escalation, conflict guards, and idempotent re-dispatch.

  • Parameters: drone_id, strategy_id, is_urgent (optional, default false)

Simulation (DTSOP)

• run_simulation — Queue a high-fidelity Digital Twin physics simulation (flood, wildfire, earthquake). Returns a job ID — subscribe to the resource URI for progress updates.

  • Parameters: scenario_id, sector_id, disaster_type, parameters, priority

• get_deployment_strategy — Generate an RL-optimized drone deployment and evacuation strategy for a confirmed incident or PDIE pre-alert.

  • Parameters: incident_id

Resources

Prompts

Example Workflows

Run a Flood Simulation

You:   "Run a flood simulation for Sector-3 with water level 4.2m"
Agent: Calls run_simulation → receives SIM-ABCD1234
       Subscribes to resq://simulations/SIM-ABCD1234
       Waits for status: completed
       Returns result URL and analysis

Full Incident Response

You:   "Validate incident INC-789 and deploy drones"
Agent: 1. Calls validate_incident(INC-789, confirmed=True)
       2. Calls get_deployment_strategy("INC-789")
       3. Reviews strategy with operator
       4. Calls update_mission_params("DRONE-Alpha", "STRAT-XYZ", urgent=True)
       5. Returns mission parameters and audit hash

Crisis Planning with Prompts

You:   Use the incident_response_plan prompt for INC-456
Agent: Receives structured template → calls tools → produces:
       - Situation Summary
       - Asset Allocation
       - Risk Assessment

Technical Architecture

C4Context
    title ResQ MCP Integration
    Person(ai, "AI Client", "Claude / VS Code / Cursor")
    System_Boundary(resq_boundary, "resq-mcp Server") {
        System(server, "resq-mcp Server", "FastMCP Interface")
        System_Boundary(backend, "ResQ Platform") {
            Component(dtsop, "DTSOP Engine", "Physics/RL Simulations")
            Component(hce, "HCE Engine", "Coordination Logic")
            Component(telemetry, "Drone Telemetry", "Real-time Status")
        }
    }
    Rel(ai, server, "Uses MCP (STDIO/SSE)")
    Rel(server, dtsop, "Executes Simulation")
    Rel(server, hce, "Validates Incidents")
    Rel(server, telemetry, "Subscribes to Data")

Module Overview

src/resq_mcp/
├── server.py              # FastMCP init, lifespan, background tasks
├── resources.py           # @mcp.resource() endpoints (drones, sims)
├── prompts.py             # @mcp.prompt() templates (incident response)
├── core/                  # Cross-cutting: config, errors, security, telemetry, timeout
├── drone/                 # Drone feed: scan, swarm, deployment (models + service)
├── dtsop/                 # Digital Twin: simulation, optimization (models + service + tools)
├── hce/                   # Hybrid Coordination: incidents, missions (models + service + tools)
└── pdie/                  # Predictive Intelligence: vulnerability, alerts (models + service)

Contributing

We use uv for dependency management and ruff for linting.

1. Setup: uv sync (installs all dependencies including dev group).
2. Test: uv run pytest
3. Lint: uv run ruff check .

Distributed under the Apache-2.0 License. Copyright 2026 ResQ.