MCP server

The IoT Pulse MCP server exposes your tenant’s fleet to any Model Context Protocol client — Claude Desktop, Cursor, or an agent you build — as a set of read/write tools and a few packaged workflow prompts. It is the second front-end of the Developer Integration Platform, alongside the CLI and the Skills pack.

Like them, it is a thin wrapper over the public API: no business logic and no new endpoints. Your tenant and scopes are resolved server-side from your API key — you never pass a tenant id, and a key can only ever touch its own tenant’s data.

Connect

The server speaks MCP over streamable HTTP at the /mcp path of your IoT Pulse MCP server URL (your administrator provides the base URL; for local development it defaults to http://localhost:3005/mcp).

Authenticate with the same scoped API key you use for the public API and CLI. Send it on every request as either header:

X-API-Key: iotp_your_key_here, or
Authorization: Bearer iotp_your_key_here.

A request with no key is rejected with 401 missing_api_key.

Most MCP clients are configured with a small JSON entry. For example:

{
  "mcpServers": {
    "iotpulse": {
      "url": "https://<your-iotpulse-mcp-host>/mcp",
      "headers": { "X-API-Key": "iotp_your_key_here" }
    }
  }
}

Mint a key in the admin app under Settings → API Keys (the secret, format iotp_…, is shown once at creation).

Tools

Every tool maps to a public API operation and requires the matching scope on your key.

Read tools

Tool	What it does
`list_devices`	List the devices in your fleet.
`get_device`	Fetch one device by id.
`list_device_groups`	List device groups (with alert config and shifts).
`list_thresholds`	List your reusable threshold packs.
`list_shifts`	List shift definitions.
`get_anomalies`	Query anomaly events (filter by device and status; paginated).

Write tools

Tool	What it does
`configure_device`	Update a device’s configuration (name, category, criticality, analysis interval, tags, location…).
`acknowledge_anomaly`	Mark an anomaly event as handled.
`update_ticket`	Update a maintenance ticket (status, severity, assignment, notes…).
`create_ticket`	Open a maintenance ticket for a device.
`set_threshold`	Create a reusable threshold pack.
`create_device_group`	Create a notification group.
`assign_shift`	Assign an existing shift to a group.

Workflow prompts

The server also exposes three packaged, multi-step workflows as MCP prompts. Any MCP client lists them with the standard prompts/list call and fetches one with prompts/get; the prompt expands into instructions that drive the tools above.

Prompt	Arguments
`onboard_device_fleet`	`groupId` (optional), `namePrefix` (optional)
`tune_thresholds_for_process_type`	`processType` (required)
`triage_anomalies_to_tickets`	`deviceId` (optional), `status` (default `open`)

The same three workflows also ship as portable Agent Skills that drive the CLI — see Skills.

Errors and limits

Because tools call the public API, they surface the same structured errors and limits: 401 invalid_api_key, 403 insufficient_scope when your key lacks a scope, 404 not_found, and per-key rate limiting (429). See the public API reference for the full table.