DOCS

MCP server

Use the HowlOps MCP server with Claude Desktop, Cursor, and other AI assistants to manage monitors, incidents, and heartbeats via the Model Context Protocol.

HowlOps supports the Model Context Protocol (MCP), letting AI assistants such as Claude Desktop and Cursor manage your monitors, incidents, and heartbeats directly from a conversation. Read-scoped API tokens enable the read tools; write-scoped tokens also unlock create / pause / resume / acknowledge / resolve.

Option A: Hosted endpoint (recommended, no install)

HowlOps serves a remote MCP endpoint at https://api.howlops.com/api/v1/mcp (or ${your-origin}/api/v1/mcp). Authenticate with a Bearer API token. Clients with native remote-MCP support (Cursor, VS Code, Codex CLI) point straight at the URL:

{
  "mcpServers": {
    "howlops": {
      "url": "https://api.howlops.com/api/v1/mcp",
      "headers": { "Authorization": "Bearer YOUR_API_TOKEN" }
    }
  }
}

Claude Desktop bridges remote HTTP MCP servers via mcp-remote:

{
  "mcpServers": {
    "howlops": {
      "command": "npx",
      "args": [
        "-y", "mcp-remote",
        "https://api.howlops.com/api/v1/mcp",
        "--header", "Authorization: Bearer YOUR_API_TOKEN"
      ]
    }
  }
}

Option B: Build the local MCP binary

Prefer running locally over stdio? The HowlOps MCP server is also a Go binary. You need Go 1.21+ installed to build it:

# Clone and build
git clone https://github.com/jsedy7/howlops-platform
cd howlops-platform/mcp-server
go build -o howlops-mcp ./cmd/main.go

# (optional) move to a directory in your PATH
mv howlops-mcp /usr/local/bin/howlops-mcp

Generate an API token

Go to Integrations → API Tokens and create a new token. Copy it — you'll need it in the next step. See Rotate an API token for token scopes.

Configure Claude Desktop (local binary)

Add the following to your claude_desktop_config.json (usually at ~/Library/Application Support/Claude/claude_desktop_config.json on macOS):

{
  "mcpServers": {
    "howlops": {
      "command": "/path/to/howlops-mcp",
      "env": {
        "HOWLOPS_API_TOKEN": "your-api-token",
        "HOWLOPS_API_BASE": "https://api.howlops.com"
      }
    }
  }
}

Configure Cursor

Add the same block to Cursor's MCP settings. Replace /path/to/howlops-mcp with the actual path to the binary you built, and your-api-token with the token you generated above.

Environment variables (local binary)

Variable	Required	Description
`HOWLOPS_API_TOKEN`	Yes	API token from Integrations → API Tokens
`HOWLOPS_API_BASE`	No	API base URL. Default: `https://api.howlops.com`

Available tools

The HowlOps MCP server exposes the following tools to your AI assistant:

Monitor management

Tool	Description
`list_monitors`	List all monitors with an optional status filter (up, down, paused)
`get_monitor`	Get monitor details by UUID
`create_monitor`	Create a new HTTP monitor (requires name, url, interval_s)
`pause_monitor`	Pause monitoring for a specific monitor
`resume_monitor`	Resume a paused monitor

Incident management

Tool	Description
`list_incidents`	List all open incidents
`get_incident`	Get incident details by UUID
`acknowledge_incident`	Acknowledge an open incident
`resolve_incident`	Resolve an incident

Heartbeats

Tool	Description
`list_heartbeats`	List all heartbeat monitors

Platform health

Tool	Description
`get_status`	Get an overall platform health summary (X up, Y down, Z paused)

Write tools (create / pause / resume / acknowledge / resolve) require a write-scoped API token. A read-scoped token can call only the read tools.

Example usage

Once configured, you can interact with HowlOps naturally:

"List all monitors that are currently down."
"Create a new monitor for https://example.com checking every 60 seconds."
"Acknowledge incident abc-123."
"Show me the platform status."
"Who is on call right now?"

Was this page helpful?