PDF

MCP and AI-Compatible Automation

Use the NetCrunch MCP Server to connect MCP-compatible tools to selected NetCrunch management operations. MCP is intended for controlled automation and assistant-assisted administration, not for unrestricted data discovery or building external monitoring systems on top of NetCrunch.

NetCrunch includes an MCP Server for integration with tools that support the Model Context Protocol.

MCP provides a standard way for AI-capable clients and automation tools to call NetCrunch management operations. It uses the same API keys, user permissions, access restrictions, and request limits as the NetCrunch REST API.

MCP does not introduce a separate security model or a separate data access layer.

The NetCrunch MCP Server is not a general query API. It does not expose NetCrunch as a database, monitoring data warehouse, topology engine, or dashboard backend. It provides access to defined management operations through an MCP-compatible interface.

When to Use MCP

Use MCP when you want to connect NetCrunch with an MCP-compatible assistant, automation client, or development tool.

Typical use cases include:

  • checking selected node properties
  • enabling or disabling monitoring for known nodes
  • updating custom fields
  • adding or removing tags
  • adding notes to nodes
  • adding or removing nodes from views
  • performing controlled maintenance operations
  • assisting operators with known administrative tasks

MCP is useful when the external tool already supports MCP and the operation can be expressed as a controlled NetCrunch management action.

Example requests:

Disable monitoring for web-server-01 for two hours.

Set the Location custom field on srv-app-03 to Warsaw Office.

Add a note to firewall-01 saying that firmware was updated.

Add node 10.10.5.20 to the Branch Office Routers view.

These examples refer to known objects and defined actions. They are appropriate for MCP.

When to Use REST API Instead

Use the REST API for deterministic integrations and repeatable automation.

REST API is usually better for:

  • scheduled scripts
  • CMDB synchronization
  • inventory updates
  • migration tools
  • repeatable maintenance workflows
  • bulk operations
  • integrations where the client knows exactly which operation to call

Example:

A nightly script updates node ownership and location fields from a CMDB export.

This is better implemented as a script using the REST API because the workflow is fixed and repeatable.

What MCP Is Not Designed For

MCP should not be used as a replacement for NetCrunch dashboards, reports, topology views, or monitoring data access.

Avoid using MCP for:

  • high-volume polling
  • broad Atlas scans
  • unrestricted data discovery
  • historical performance data extraction
  • rebuilding NetCrunch views in an external system
  • creating an external monitoring portal
  • querying all monitoring data as if NetCrunch were a database
  • replacing NetCrunch UI workflows with a separate product layer

If a workflow requires large-scale data extraction or reporting, use dedicated NetCrunch reports, exports, UI views, or purpose-built integrations.

MCP and AI Assistants

An AI assistant connected through MCP can help with selected administrative tasks.

For read-only use, an assistant can help inspect known objects and explain returned information. For write-capable use, the assistant can prepare or execute controlled changes using the permissions of the API key.

Recommended read-only tasks:

  • show selected node properties
  • check monitoring state
  • review custom field values
  • list tags assigned to a node
  • check whether a node belongs to a view
  • summarize returned information for an operator

Recommended write-capable tasks:

  • disable monitoring for approved maintenance
  • re-enable monitoring after maintenance
  • update selected custom fields
  • add or remove tags
  • add a node note
  • add or remove a node from a view

For write-capable assistants, start with narrow permissions. Use confirmation before applying changes that affect monitoring state, node configuration, view membership, or multiple objects.

Controlled Automation Pattern

For write operations, MCP clients should follow a controlled workflow:

  1. Identify the target object.
  2. Read the current state when needed.
  3. Prepare the requested change.
  4. Show the planned action to the user.
  5. Ask for confirmation when the change is not read-only.
  6. Execute the operation.
  7. Read back the result when practical.
  8. Report the final state.

Example:

User:
Disable monitoring for web-server-01 until 18:00 and add a maintenance note.

MCP client:
1. Reads the current monitoring state of web-server-01
2. Prepares the disable-monitoring operation
3. Adds the maintenance note
4. Executes the change after confirmation
5. Reads back the monitoring state
6. Reports the result

This pattern reduces accidental changes and keeps the operator in control.

Synchronization Use Cases

MCP can support synchronization workflows when the external system provides known objects and explicit values.

Examples:

  • update node custom fields from a CMDB
  • apply tags from an asset inventory
  • add known nodes to views based on location or ownership
  • add notes from an external ticketing system
  • disable monitoring during approved maintenance windows

Synchronization should use stable identifiers whenever possible, such as:

  • node ID
  • node name
  • IP address
  • DNS name
  • view path
  • policy name where applicable

Avoid open-ended synchronization workflows that attempt to discover and reconstruct the whole NetCrunch configuration externally.

Permissions

MCP uses NetCrunch API keys.

The API key determines:

  • which NetCrunch user context is used
  • which objects are visible
  • which operations are allowed
  • whether write operations are permitted
  • whether the key is restricted by source address
  • whether the key has expired

Recommended practices:

  • create a dedicated user for MCP access
  • grant only the permissions required by the integration
  • use read-only API keys for inspection-only scenarios
  • avoid using administrator accounts
  • restrict API keys by source address where possible
  • set an expiration date for temporary integrations
  • review activity after enabling MCP access

Do not connect an unrestricted administrator API key to an experimental AI assistant or external automation tool.

Rate Limiting

MCP requests are rate-limited to protect the NetCrunch Server from excessive tool calls.

AI-capable clients may generate several tool calls for a single user request. Rate limiting helps prevent accidental loops, inefficient workflows, and excessive load on the server.

When an MCP client reaches the request limit, reduce the number of calls or add throttling on the client side.

Recommended improvements:

  • avoid repeated reads of the same object
  • cache known object identifiers
  • operate on selected nodes or views
  • avoid broad scans
  • batch predictable work where possible
  • retry with backoff after rate-limit errors

Security Guidelines

Before enabling MCP access:

  • decide what the MCP client is allowed to do
  • create a dedicated API key
  • start with read-only access when possible
  • restrict access to required views, nodes, or organizations
  • avoid shared administrator credentials
  • keep rate limiting enabled
  • require confirmation for write operations
  • review API activity during initial use

MCP clients should be treated like external API clients. They can simplify work, but they should not receive broader access than the user or workflow requires.

REST API or MCP

Use REST API when:

  • the workflow is fixed
  • the task is scheduled
  • the client is a script or service
  • exact repeatability matters
  • the integration performs bulk synchronization
  • the developer knows which endpoint to call

Use MCP when:

  • the client already supports MCP
  • the workflow is operator-assisted
  • the task is interactive
  • the operation uses known NetCrunch objects
  • the assistant should help prepare or explain selected actions
  • the integration benefits from a standard AI-tool interface

REST API remains the preferred option for deterministic automation.

MCP is an additional compatibility layer for MCP-capable clients and assistant-assisted administration.

Summary

The NetCrunch MCP Server provides AI-compatible access to selected NetCrunch management operations.

It is intended for controlled automation, synchronization, and assistant-assisted administration. It is not intended to expose NetCrunch as a general monitoring engine, database, reporting layer, or external product platform.

Use MCP for practical administrative actions. Use REST API for repeatable integrations. Use NetCrunch UI, dashboards, reports, and future built-in AI features for deeper product-aware analysis and monitoring workflows.

ai assistantapi keyautomationcustom fieldsmaintenance modemcp servermodel context protocolnetcrunchnotespermissionsrate limitingrest apisynchronizationtags