Skip to main content
AI coding agents — like those in Cursor, Claude Desktop, and Windsurf — can create monitors, investigate incidents, and deploy configurations through tool APIs. This article explains the patterns and possibilities.

​
What AI agents can do with monitoring

​
Create and manage monitors

Instead of navigating a dashboard, describe what you want:
“Create an HTTP monitor for api.example.com/health that checks every 60 seconds from US and EU regions, asserts a 200 status code, and alerts our Slack channel on failure.”
The agent translates this into the right API calls, creating the monitor with all specified configuration.

​
Investigate incidents

When something breaks, an agent can:
  1. List active incidents
  2. Pull recent check results for the affected monitor
  3. Check dependency status for upstream services
  4. Correlate timing with recent deployments
  5. Suggest probable causes based on the data

​
Generate and maintain configs

Agents can read your devhelm.yml, understand the current monitoring setup, and suggest changes:
  • Add monitors for new endpoints
  • Update assertions when API contracts change
  • Adjust frequencies based on traffic patterns

​
Produce reports

Ask for a summary and the agent pulls the data:
“Show me the uptime report for all production monitors over the last 30 days.”

​
How agents connect to monitoring tools

​
Direct API calls

Agents can call REST APIs directly using HTTP tools: 
POST /api/v1/monitors
Authorization: Bearer <token>
Content-Type: application/json

{ "name": "...", "type": "HTTP", ... }
This works but requires the agent to know the full API schema.

​
Model Context Protocol (MCP)

MCP is a standardized protocol that exposes tool catalogs to AI agents. Instead of raw API knowledge, the agent discovers available tools: 
Tools available:
  - create_monitor(name, type, url, frequency, ...)
  - list_incidents(status, limit)
  - get_check_results(monitorId, limit)
  ...
The agent picks the right tool, fills in parameters, and executes. MCP handles authentication, serialization, and error handling.

​
CLI as a tool

Some agents can execute CLI commands: 
devhelm monitors create \
  --name "API Health" \
  --type HTTP \
  --url https://api.example.com/health \
  --frequency 60
This works well for agents that already have terminal access.

​
Practical use cases

Use caseAgent behavior
New service deploymentCreate monitors + alerts for all endpoints
Incident investigationPull results, check dependencies, suggest causes
Config reviewAudit existing monitors for coverage gaps
Compliance checkVerify all production services have monitors
Environment setupReplicate monitoring config across staging/prod

​
Limitations and considerations

  • Agents make mistakes — always review monitor configurations before deploying
  • Context window limits — agents may lose track of complex multi-step operations
  • Authentication scope — grant agents the minimum permissions needed
  • Audit trail — API calls from agents should be traceable (use dedicated API keys)

​
DevHelm MCP Server

DevHelm’s MCP server exposes 67 tools covering monitors, incidents, alerts, deployments, and more.

MCP Server setup

Connect your AI agent to DevHelm.

MCP protocol overview

How MCP works under the hood.