PRTG MCP Server
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@PRTG MCP Server are there any sensors in alarm state right now?"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
PRTG MCP Server
MCP server for Paessler PRTG's REST API v2 - device/sensor status, monitoring data (channel measurement history), and alerts (sensors in an alarm state) across a PRTG network/infrastructure monitoring instance - for AI assistants and the WYRE Conduit gateway.
Authentication
PRTG authenticates with a static Bearer API key, generated in the PRTG web UI under Account Settings -> My Account -> API Keys. No admin approval is required to generate one - any PRTG user can create a key for themselves. Unlike most connectors in this fleet, PRTG has no fixed hosted API endpoint: this connector supports self-hosted PRTG Network Monitor at its own base URL (PRTG Hosted Monitor does not provide API v2 yet; see below), so it requires both an API key and a base URL. The key is sent as Authorization: Bearer <key> to <base URL>/api/v2/.... In gateway mode both arrive per-request via the X-PRTG-Api-Key and X-PRTG-Base-URL headers; in local/stdio mode they're read once from PRTG_API_KEY and PRTG_BASE_URL.
Credential scope
Vendor-documented, not independently verified against a live account (see Verification below for why): per Paessler's own PRTG API v2 documentation, an API key inherits the full permissions of the PRTG user account that created it - keys cannot be independently scoped to read-only. Paessler's own guidance for AI-assistant integrations is explicit: "grant [API keys used in AI assistants] the lowest user role that fits the assistant's task," and recommends creating a dedicated read-only PRTG user account first, then generating the API key under that account, rather than relying on the key itself to restrict access. This connector cannot enforce that on PRTG's side - it only implements the read-only tool surface described below - so the actual read-only guarantee depends on the customer following Paessler's own recommendation when they generate the key.
Base URL (on-premises)
On PRTG Network Monitor (on-premises), API v2 is served only by the PRTG Application Server, not by the classic web interface. Use the Application Server's HTTPS URL, by default port 1616 - e.g. https://prtg.example.com:1616. The Application Server also listens for plain HTTP on port 1615, but every request carries the bearer API key, so use http://...:1615 only on a trusted network. The classic web interface URL (port 80/443) does not serve /api/v2 and typically returns 404, a redirect, or the HTML login page, which looks like an authentication failure.
The Application Server exists only after New UI and API v2 has been activated: in the classic UI go to Setup | Overview and select Activate (ports 1616/1615 and 23580 must be free; the core server restarts). Per Paessler, it is activated by default for new PRTG users as of PRTG 25.2.106. Existing installs, upgrades, and any install where Setup | Overview still shows Activate New UI And API v2 must activate it manually. Paessler also states that API v2 is not yet available on PRTG Hosted Monitor or on PRTG installs running in cluster mode. See Paessler's I want to use the new UI and API v2. What do I need to know? and the New User Interface manual page.
Verification
This connector was built directly against PRTG's official, current OpenAPI v2 specification (prtg.api.yaml, published at paessler.com/support/prtg/api/v2/oas/), not against secondary documentation or a naming convention - every tool below maps to one real, named operation in that spec, with its exact query parameters. What it is not is independently verified against a live PRTG instance: PRTG's free tiers are (a) a 30-day trial of PRTG Network Monitor, which is Windows-only on-premises software, and (b) a 10-day trial of PRTG Hosted Monitor (app.my-prtg.com/signup), a cloud SaaS signup whose verification flow could not be completed in this build environment (no Windows host to install PRTG Network Monitor on; no email inbox available to complete PRTG Hosted Monitor's signup confirmation). Both are genuinely self-serve for a human with a Windows machine or an email inbox to check - this is an environment limitation of the build, not a vendor-side approval gate.
Related MCP server: GetMonitor MCP Server
Configuration
Env var | Description |
| Bearer API key issued by the PRTG web UI. |
| Base URL of the PRTG Application Server (which serves API v2), e.g. |
|
|
|
|
| When set, the HTTP transport requires a valid |
|
|
Request limits
HTTP request bodies are capped at 256 KiB (262144 bytes). After the
X-Gateway-S2Scheck passes, aContent-Lengthabove the cap returns 413 with a JSON-RPC error; bodies withoutContent-Length(e.g. chunked) are capped by the MCP SDK transport, which also returns 413. Oversized tool input returns the fixed errorTool input exceeds the maximum of 10000 elements. Unauthenticated requests still get 401 first.tools/callarguments are capped at 10,000 elements (every array item and object key at any depth). Larger inputs return anisErrortool result and are never dispatched.
These limits address CVE-2026-111768 / GHSA-46jr-4wx8-v453 (memory exhaustion during validation) and GHSA-f897-62hf-6gxq (no request body / batch cap).
Tools
Devices
prtg_list_devices- list monitored devices (hostname/IP, status, parent group/probe).prtg_get_device- get the metrics and settings of a single device.
Groups
prtg_list_groups- list groups (the organizational folders that contain devices).prtg_get_group- get the metrics and settings of a single group.
Probes
prtg_list_probes- list probes (the monitoring agents - local, remote, or mini - that run the actual sensor checks).prtg_get_probe- get the metrics and settings of a single probe.prtg_get_probe_network_info- get network information for a probe.
Sensors
prtg_list_sensors- list sensors (the individual checks - ping, HTTP, SNMP, etc. - attached to a device), with their current status.prtg_get_sensor- get the metrics and settings of a single sensor.prtg_get_sensor_data- get channel measurement history (the actual monitoring/performance data) for a sensor.prtg_list_alarms- list sensors currently in a Down/Warning/Unusual/etc. alarm state. PRTG's alerting surface.
Sensor status summary
prtg_get_sensor_status_summary- get a summary of all sensor states across the whole instance.prtg_get_object_sensor_summary- get a sensor-state summary scoped to one probe, group, or device.
Channels
prtg_get_channel- get the metrics and settings (units, limits, current value) of a single channel on a sensor.
System
prtg_get_version- get the PRTG core/appserver version. Useful as a credential/connectivity sanity check.
Scope
This is a deliberately narrow, read-only v1 surface covering exactly device/sensor status, monitoring data, and alerts - nothing else. PRTG API v2 documents a large surface (object CRUD, autodiscovery, user/API-key administration, AI-assistant endpoints, an entire /experimental namespace); this connector implements 15 GET operations from it (10 stable, non-experimental, and 5 documented-but-deprecated - see below), verified one-by-one against the official OpenAPI spec, and excludes every write/pause/resume/scan/acknowledge/discovery/administration operation, plus every genuinely /experimental/* endpoint, by design, not by oversight.
A few of the 15 implemented endpoints (GET /devices, /groups, /probes, /sensors, /sensors/alarms) are marked deprecated in PRTG's own v2 spec, in favor of an /experimental/* replacement. This connector deliberately uses the documented, currently-working deprecated endpoint rather than its /experimental replacement: PRTG's own spec states experimental endpoints "might change," while a deprecated endpoint "still works" and is the only non-experimental way to list PRTG's core object types at all. A stable-but-deprecated endpoint was judged the safer choice for a production connector over an unstable-but-current one; every /experimental/* endpoint (list or otherwise) is excluded outright below.
Hard-excluded (mutation - pause/resume/scan) - never implemented: POST /devices/{id}/pause, /devices/{id}/resume, /devices/{id}/scan, /devices/pause, /devices/resume, /devices/scan, /groups/{id}/pause, /groups/{id}/resume, /groups/{id}/scan, /groups/pause, /groups/resume, /groups/scan, /probes/{id}/pause, /probes/{id}/resume, /probes/{id}/scan, /probes/pause, /probes/resume, /probes/scan, /sensors/{id}/pause, /sensors/{id}/resume, /sensors/{id}/scan, /sensors/pause, /sensors/resume, /sensors/scan, /users/{id}/pause, /users/{id}/resume, /users/pause, /users/resume.
Hard-excluded (acknowledge-alert) - never implemented: POST /sensors/{id}/acknowledge, /sensors/acknowledge - acknowledging an alarm is a write action on PRTG's alerting state, excluded per this connector's hard scope boundary even though it reads like a status query.
Hard-excluded (object mutation) - never implemented: POST /objects/{id}/move, /objects/{id}/clone, and the entire /experimental namespace's create/update/delete operations (PATCH /experimental/groups/{id}, /experimental/probes/{id}, /experimental/sensors/{id}; POST /experimental/groups/{id}/group, /experimental/probes/{id}/group, /experimental/groups/{id}/device, /experimental/probes/{id}/device, /experimental/devices/{id}/sensor; DELETE /experimental/devices/{id}) and its experimental read-only list mirrors (GET /experimental/devices, /experimental/groups, /experimental/probes, /experimental/sensors, /experimental/channels, /experimental/objects, /experimental/devices/templates) - excluded as a whole namespace PRTG's own docs describe as subject to change, not individually vetted as safe.
Hard-excluded (autodiscovery) - never implemented: GET /autodiscoveries, /experimental/autodiscoverytasks, /experimental/autodiscoverytasks/{id}; POST /experimental/devices/{id}/autodiscovery, /experimental/groups/{id}/autodiscovery, /experimental/devices/{id}/metascan, /experimental/probes/{id}/autodiscoverytask, /experimental/autodiscoverytasks/{id}/start - autodiscovery inventories and triggers network scans; out of this connector's device/sensor/monitoring/alert scope.
Hard-excluded (user/API-key/session/license administration - identity and credential management, not device/sensor/alert data) - never implemented: GET/POST/PATCH /users*, /experimental/users*, /usergroups*, /experimental/usergroups*, /users/api-keys/find, /users/{id}/api-keys, DELETE /users/api-keys/{id}, POST/GET/DELETE /session (login/renew/logout), POST /users/request-password, /users/reset-password, GET /experimental/license.
Hard-excluded (object schema/generic-CRUD/settings/AI-feature/misc administration) - out of scope, never implemented: GET/POST/PATCH /schemas/{kind}, /experimental/schemas/*, GET /experimental/objects, /objects, /objects/count, /libraries, /lookup-definitions*, /setting-lookups/{name}, /settings/public, /health (unauthenticated liveness only - prtg_get_version is this connector's authenticated sanity check instead), /devices/icons, /experimental/feature-toggles, /experimental/featureflags, /experimental/ai/analyze, /experimental/ai/status, /experimental/timeseries/{id}, /experimental/timeseries/{id}/{type}, /channels (list) and /channels/data (both deprecated in favor of the implemented GET /channels/{id}, itself not deprecated), /channels/{id}/overview, /devices/{id}/overview, /groups/{id}/overview, /probes/{id}/overview, /sensors/{id}/overview (all deprecated in favor of the implemented non-overview GET .../{id} equivalents).
They can be added as a follow-up if there's demand, after a deliberate scope decision - not by default.
Development
npm install
npm run build
npm test
npm run lint # tsc --noEmitDocker
docker build -t prtg-mcp .
docker run -p 8080:8080 -e PRTG_API_KEY=... -e PRTG_BASE_URL=https://prtg.example.com:1616 prtg-mcpAvailable Tools
15 toolsprtg_get_channelA
Get the metrics and settings (units, limits, current value) of a single channel on a sensor by ID. Channel IDs appear in prtg_get_sensor_data and prtg_get_sensor results.
Inputs: id (required), include.
prtg_get_deviceB
Get the metrics and settings of a single device by ID.
Inputs: id (required), include.
prtg_get_groupA
Get the metrics and settings of a single group by ID.
Inputs: id (required), include.
prtg_get_object_sensor_summaryA
Get a sensor-state summary scoped to one probe, group, or device.
Inputs: id (required).
prtg_get_probeA
Get the metrics and settings of a single probe by ID.
Inputs: id (required), include.
prtg_get_probe_network_infoB
Get network information (the probe host's own connectivity details) for a probe.
Inputs: id (required).
prtg_get_sensorB
Get the metrics and settings of a single sensor by ID.
Inputs: id (required), include.
prtg_get_sensor_dataA
Get channel measurement history (the actual monitoring/performance data) for a sensor.
Inputs: id (required), limit, filter, offset, +1 more.
prtg_get_sensor_status_summaryA
Get a summary of all sensor states (counts by status - Up, Down, Warning, Paused, etc.) across the whole PRTG instance.
Inputs: no named parameters.
prtg_get_versionA
Get the PRTG core server and appserver version. Useful as a quick credential/connectivity sanity check.
Inputs: no named parameters.
prtg_list_alarmsA
List sensors currently in a Down/Warning/Unusual/etc. alarm state that match a filter. PRTG's alerting surface - use this to find what's currently wrong.
Inputs: limit, filter, offset, sort_by, +1 more.
prtg_list_devicesA
List monitored devices (hostname/IP, status, parent group/probe). Each device's id chains into prtg_get_device, prtg_list_sensors (filter by parent), and prtg_get_object_sensor_summary.
Inputs: limit, filter, offset, sort_by, +1 more.
prtg_list_groupsA
List groups (the organizational folders that contain devices). Each group's id chains into prtg_get_group and prtg_get_object_sensor_summary.
Inputs: limit, filter, offset, sort_by.
prtg_list_probesA
List probes (the monitoring agents - local, remote, or mini - that run the actual sensor checks). Each probe's id chains into prtg_get_probe, prtg_get_probe_network_info, and prtg_get_object_sensor_summary.
Inputs: limit, filter, offset, sort_by.
prtg_list_sensorsA
List sensors (the individual checks - ping, HTTP, SNMP, etc. - attached to a device), with their current status. Each sensor's id chains into prtg_get_sensor and prtg_get_sensor_data.
Inputs: limit, filter, offset, sort_by.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 15 tools
Each tool maps cleanly to a distinct PRTG object type or action: probes, groups, devices, sensors, channels, alarms, and version. The hierarchical relationships are explicit in the descriptions, so there is little risk of selecting the wrong tool.
All tools follow a consistent `prtg_` + verb + noun pattern, with `list` for collections and `get` for single objects or details. Naming style is uniform snake_case throughout, making the API surface predictable.
Fifteen tools is at the upper edge of the ideal range but each one maps to a necessary operation in the PRTG hierarchy. The count is well-scoped for a monitoring-focused server with no redundant or filler tools.
The read-only monitoring workflow is well covered: navigate probes/groups/devices, list sensors, retrieve channel data, and inspect alarms/status summaries. The main gap is the lack of operational actions like pausing sensors or acknowledging alarms, but those may be intentionally outside the server's scope.
Maintenance
Related MCP Connectors
Lets AI agents analyze API metrics, request logs, and traces, and manage apps and alerts.
Read-only access to your MetaTrader 4 and 5 accounts for AI agents: balances, open positions, trade history with win rate, live quotes, symbol specs and candles. It cannot place, modify or close orders.
Read-only search of 2,189 AI agent tools (MCP servers, x402 APIs) with daily automated trust scores.
Read-only access to your Infrapage dashboards: pages, widgets, live values, weekly recaps.
Related MCP Servers
- AlicenseNot gradedqualityDmaintenanceEnables AI assistants to query Prometheus metrics, monitor alerts, and analyze system health through read-only access to your Prometheus server with built-in query safety and optional AI-powered metric analysis.MIT

GetMonitor MCP Serverofficial
AlicenseCqualityCmaintenanceConnects AI assistants to GetMonitor status pages, monitors, incidents, and maintenance schedules via read-only tools.10015 npmApache 2.0- AlicenseAqualityBmaintenanceEnables read-only access to AlertOps incident alerting and on-call data, including alerts, users, groups, integrations, and schedules, for AI assistants.11Apache 2.0
- AlicenseBqualityBmaintenanceEnables AI assistants to query and manage Zabbix monitoring data—hosts, items, triggers, problems, and health checks—and generate daily, weekly, and monthly operations reports.21GPL 3.0