Prey MCP server

A Model Context Protocol (MCP) server for the Prey Public API.

This server provides read and write access to your Prey account with guardrails.

Requirements

• Go 1.22+
• A Prey API Key with appropriate permissions

Features

Read-only tools (default):

• Account summary
• Users list and details
• Devices list and details
• Device reports list and details
• Device location history (JSON or CSV)
• Labels list and details
• Zones list and details
• Automations list and details
• Mass actions list and details

Write tools (opt-in):

• Trigger device action (alarm/alert/lock)
• Set device missing/recovered
• Create label
• Create/update zone
• Delete device

Configuration

Environment variables:

• PREY_API_KEY (required)
• PREY_API_BASE (default: https://api.preyproject.com/v1)
• PREY_TIMEOUT_MS (default: 30000)
• PREY_ALLOW_WRITE (default: false)
• PREY_ALLOWED_TOOLS (comma-separated allowlist)
• PREY_DEBUG (default: false)
• PREY_RATE_LIMIT_DISABLE (default: false)

Optional per-request headers (multi-tenant scenarios):

• X-Prey-URL
• X-Prey-API-Key

Rate limiting

By default the client enforces Prey limits (per API key):

• 2 requests/second
• 60 requests/minute
• 10,000 requests/hour

Disable with PREY_RATE_LIMIT_DISABLE=true.

Tools

• prey.account.get
• prey.users.list
• prey.users.get
• prey.devices.list
• prey.devices.get
• prey.devices.delete
• prey.devices.reports.list
• prey.devices.reports.get
• prey.devices.location_history.get
• prey.labels.list
• prey.labels.get
• prey.labels.create
• prey.zones.list
• prey.zones.get
• prey.zones.create
• prey.zones.update
• prey.automations.list
• prey.automations.get
• prey.mass_actions.list
• prey.mass_actions.get
• prey.devices.action.trigger
• prey.devices.status.set

Transport

Supported transports:

• stdio
• sse
• streamable-http

Examples

Docker (streamable-http):

docker run --rm -p 8000:8000 \
  -e PREY_API_KEY=YOUR_KEY \
  preyproject/mcp-prey:latest \
  --transport streamable-http --address 0.0.0.0:8000 --endpoint-path /mcp

Docker (SSE):

docker run --rm -p 8000:8000 \
  -e PREY_API_KEY=YOUR_KEY \
  preyproject/mcp-prey:latest \
  --transport sse --address 0.0.0.0:8000 --base-path /

Docker (write enabled):

docker run --rm -p 8000:8000 \
  -e PREY_API_KEY=YOUR_KEY \
  -e PREY_ALLOW_WRITE=true \
  preyproject/mcp-prey:latest \
  --transport streamable-http --address 0.0.0.0:8000 --endpoint-path /mcp

Stdio:

PREY_API_KEY=... \
PREY_ALLOW_WRITE=false \
./mcp-prey --transport stdio

SSE:

PREY_API_KEY=... \
./mcp-prey --transport sse --address localhost:8000 --base-path /

Streamable HTTP:

PREY_API_KEY=... \
./mcp-prey --transport streamable-http --address localhost:8000 --endpoint-path /mcp

Claude Desktop (example config):

{
  "mcpServers": {
    "prey": {
      "command": "docker",
      "args": [
        "run",
        "--rm",
        "-p",
        "8000:8000",
        "-e",
        "PREY_API_KEY=YOUR_KEY",
        "preyproject/mcp-prey:latest",
        "--transport",
        "streamable-http",
        "--address",
        "0.0.0.0:8000",
        "--endpoint-path",
        "/mcp"
      ]
    }
  }
}

ChatGPT Developer Mode (SSE or streamable HTTP):

1. Enable Developer mode in ChatGPT settings (Connectors → Advanced → Developer mode).
2. Add a connector with your MCP server URL.
3. Use it in a chat via the Developer mode tool picker.

Codex CLI (example):

codex mcp add prey --url http://localhost:8000/mcp
codex mcp list

Codex config (example ~/.codex/config.toml):

[mcp_servers.prey]
url = "http://localhost:8000/mcp"

Development

Start the dev container:

task mcp-prey:dev:up

Run tests (inside container):

task mcp-prey:dev:exec SV=mcp-prey-dev CMD='go test ./...'

Notes

• Write tools are disabled unless PREY_ALLOW_WRITE=true.
• For large fleets, use pagination; page_size is capped at 100.
• CSV location history is returned as base64 with content_type.

License

MIT