MCP Integration
Connect any MCP-compatible AI agent to your PulseGuard monitors — read uptime stats, inspect incidents, and manage monitors through natural language.
Prerequisites
- A PulseGuard account with at least one monitor configured.
- A PulseGuard API token — generate one under Profile → API Keys. Use a Read-only token for telemetry queries, or a Full Access token to create, update, pause, configure, and delete monitors on paid plans.
-
Node.js 18+ — required if you use the
mcp-remotebridge for Claude Desktop. Download from nodejs.org. Verify withnode -v. Not needed if you use the Connectors UI or Cursor/Windsurf. - An MCP-compatible client. Supported clients:
| Client | Min version | Notes |
|---|---|---|
| Claude Desktop | 0.7.0+ | Connectors UI (OAuth) or mcp-remote (Bearer token) |
| Cursor | 0.43.0+ | Settings → MCP → Add server |
| Windsurf | 1.0.0+ | Settings → Cascade → MCP |
| Antigravity | any | Edit ~/.gemini/antigravity/mcp_config.json |
| Any SSE client | MCP spec 2024-11-05+ | Pass Authorization: Bearer header |
Token Types
PulseGuard uses Passport API tokens and OAuth access tokens. Two permission levels are available:
| Token Type | List / Read monitors | Read stats & incidents | Create monitors | Pause / Resume | Delete via AI |
|---|---|---|---|---|---|
| Read-only | ✓ | ✓ | ✗ | ✗ | ✗ |
| Full Access | ✓ | ✗ | ✓ | ✓ | ✓ |
Full Access tokens carry the mcp:full ability. Read-only tokens carry the read ability. Full Access write actions require an eligible paid plan; monitor deletion also requires the exact monitor ID.
Installation
The MCP server endpoint is:
https://pulseguard.saniolabs.com/mcp
No config file or API key needed — Claude Desktop handles the OAuth flow automatically.
- 1 Open Claude Desktop → Settings → Integrations.
- 2 Click Add Integration and paste the MCP server URL:
https://pulseguard.saniolabs.com/mcp
- 3 Claude opens a browser window — log in to PulseGuard and click Authorize.
-
4
The integration appears in Claude Desktop. Ask
"List my PulseGuard monitors"to verify.
Uses the mcp-remote bridge to connect Claude Desktop via a Bearer token. Requires Node.js 18+.
Config file path
{ "mcpServers": { "pulseguard": { "command": "npx", "args": [ "mcp-remote@latest", "https://pulseguard.saniolabs.com/mcp", "--header", "Authorization: Bearer YOUR_API_TOKEN" ] } } }
YOUR_API_TOKEN with a token from Profile → API Keys. Fully quit Claude Desktop (Cmd+Q / Alt+F4), then reopen.
Verify: ask Claude "List my PulseGuard monitors" — it should call list-monitors-tool and return your monitors.
Open or create ~/.cursor/mcp.json (global) or .cursor/mcp.json (per-project):
Config file path
{ "mcpServers": { "pulseguard": { "url": "https://pulseguard.saniolabs.com/mcp", "headers": { "Authorization": "Bearer YOUR_API_TOKEN" } } } }
Verify: open Cursor Chat and type "List my PulseGuard monitors" — Cursor should invoke list-monitors-tool.
Open or create ~/.codeium/windsurf/mcp_config.json:
Config file path
{ "mcpServers": { "pulseguard": { "serverUrl": "https://pulseguard.saniolabs.com/mcp", "headers": { "Authorization": "Bearer YOUR_API_TOKEN" } } } }
Cmd/Ctrl+Shift+P → Reload Window).
Verify: in Cascade ask "List my PulseGuard monitors".
Open or create ~/.gemini/antigravity/mcp_config.json and add the pulseguard entry:
Config file path
"pulseguard": { "serverUrl": "https://pulseguard.saniolabs.com/mcp", "headers": { "Authorization": "Bearer YOUR_API_TOKEN" } }
YOUR_API_TOKEN with a token from Profile → API Keys.
Verify: ask Antigravity "List my PulseGuard monitors" — it should call list-monitors-tool and return your monitors.
Any MCP client that supports Streamable HTTP transport works. The server accepts JSON-RPC 2.0 POST requests with a single Authorization header.
# curl — verify the Streamable HTTP endpoint curl -s -X POST \ -H "Authorization: Bearer YOUR_API_TOKEN" \ -H "Content-Type: application/json" \ -d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2025-03-26","capabilities":{},"clientInfo":{"name":"curl-test","version":"1.0"}}}' \ https://pulseguard.saniolabs.com/mcp
A successful response returns a JSON-RPC result with serverInfo and capabilities. From there the MCP handshake is complete and tools are available.
Available Tools
PulseGuard exposes read tools for monitor health, incidents, SSL, PageSpeed, status pages, charts, and notification channels. Paid plans can also create, edit, pause, resume, configure, and delete monitors safely.
list-monitors-tool
Read
get-uptime-stats-tool
Read
get-monitor-incidents-tool
Read
list-all-incidents-tool
Read
create-monitor-tool
Full Access
update-monitor-tool
Full Access
toggle-monitor-pause-tool
Full Access
delete-monitor-tool
Full Access
list-monitors-tool
Read tokenReturns all monitors configured in the authenticated user's account.
Input schema
{ "type": "object", "properties": {} }
No input required.
Example response
[
{
"id": "a1b2c3d4-e5f6-7890-abcd-ef1234567890",
"name": "Production API",
"url": "https://api.example.com",
"status": "up",
"type": "http",
"check_interval": 300,
"paused_at": null
}
]
get-uptime-stats-tool
Read tokenReturns uptime percentage and average response time for a given time period. Use period to choose between "24h" or "30d" (default).
Input schema
{ "type": "object", "required": ["monitor_id"], "properties": { "monitor_id": { "type": "string", "description": "UUID of the monitor" }, "period": { "type": "string", "description": "\"24h\" or \"30d\" (default: \"30d\")" } } }
Example response
{
"monitor_name": "Production API",
"status": "up",
"period": "last 30 days",
"uptime_percentage": "99.85%",
"average_response_time_ms": 142,
"total_pings_analyzed": 129600, // 30d × 24h × 60min × 3 pings/min
"total_incidents": 3
}
get-monitor-incidents-tool
Read tokenReturns the 10 most recent incidents for a monitor, ordered by started_at descending.
Input schema
{ "type": "object", "required": ["monitor_id"], "properties": { "monitor_id": { "type": "string", "description": "UUID of the monitor" } } }
Example response
[
{
"id": 12,
"monitor_id": "a1b2c3d4-...",
"started_at": "2026-03-28T14:22:00.000000Z",
"resolved_at": "2026-03-28T14:35:00.000000Z",
"duration": 780,
"root_cause": "Connection timeout"
}
]
create-monitor-tool
Full Access tokenCreates a new HTTP or keyword monitor. Returns the created monitor object. Fails if the account's monitor limit is reached.
Input schema
{ "type": "object", "required": ["name", "url"], "properties": { "name": { "type": "string", "description": "Display name, e.g. My API" }, "url": { "type": "string", "description": "Full URL, e.g. https://api.example.com" }, "type": { "type": "string", "enum": ["http", "keyword"], "description": "Default: http" }, "check_interval": { "type": "integer", "description": "Seconds between checks. Default: 300" }, "timeout": { "type": "integer", "description": "Seconds before marking down. Default: 10" }, "monitor_ssl": { "type": "boolean", "description": "Monitor SSL cert. Default: false" }, "notify_uptime": { "type": "boolean", "description": "Email alerts. Default: true" }, "monitor_domain": { "type": "boolean", "description": "Monitor domain expiry. Default: false" } } }
Example response
Monitor created successfully.
{
"id": "f7e6d5c4-b3a2-1098-fedc-ba9876543210",
"name": "Staging API",
"url": "https://staging.example.com",
"type": "http",
"status": "pending",
"check_interval": 300,
"timeout": 10,
"monitor_ssl": false,
"notify_uptime": true,
"monitor_domain": false
}
toggle-monitor-pause-tool
Full Access tokenPauses or resumes a monitor. Pass pause: true to stop checks, pause: false to resume. Paused monitors show status "paused"; resumed monitors revert to "pending" until the next check.
Input schema
{ "type": "object", "required": ["monitor_id", "pause"], "properties": { "monitor_id": { "type": "string", "description": "UUID of the monitor to modify" }, "pause": { "type": "boolean", "description": "true to pause, false to resume" } } }
Example response (pause)
Monitor paused successfully.
Example response (resume)
Monitor unpaused successfully.
list-all-incidents-tool
Read tokenReturns recent incidents across all of the user's monitors in a single call. Incidents are sorted by started_at descending and include monitor_name for context.
Input schema
{ "type": "object", "properties": { "limit": { "type": "integer", "description": "Max incidents per monitor (1–50). Default: 10" } } }
All parameters optional. Omit limit to use the default of 10.
Example response
[
{
"monitor_name": "Production API",
"monitor_id": "a1b2c3d4-...",
"incident_id": 42,
"started_at": "2026-03-28T14:22:00.000000Z",
"resolved_at": "2026-03-28T14:35:00.000000Z",
"duration": "13m 0s",
"root_cause": "Connection timeout"
},
{
"monitor_name": "Staging API",
"monitor_id": "b2c3d4e5-...",
"incident_id": 38,
"started_at": "2026-03-27T09:10:00.000000Z",
"resolved_at": "2026-03-27T09:11:30.000000Z",
"duration": "1m 30s",
"root_cause": "Gateway Timeout"
}
]
Example Prompts
Copy these directly into your AI agent. The agent will call the appropriate tools automatically.
List all my PulseGuard monitors and show me their current status and check intervals.
What is the uptime percentage and average response time for my Production API monitor? Give me the monitor ID too.
Fetch the last 10 incidents for monitor ID a1b2c3d4-e5f6-7890-abcd-ef1234567890 and tell me if there is a recurring failure pattern based on root_cause and timing.
List all my monitors, then fetch uptime stats for each one and rank them by average response time from fastest to slowest.
Create three HTTP monitors: "API Gateway" at https://api.example.com, "Auth Service" at https://auth.example.com, and "Dashboard" at https://app.example.com. Set check_interval to 60 for all three and enable SSL monitoring.
Troubleshooting
401 Authentication error — "Unauthenticated" or HTTP 401
The Authorization header is missing or malformed.
- Confirm the token value starts with no extra whitespace.
- The header must be exactly
Authorization: Bearer <token>. - Regenerate the token in Profile → API Keys if unsure.
403 "does not have full write permissions" error
You are using a Read-only token and calling a write tool (create-monitor-tool or toggle-monitor-pause-tool).
Generate a Full Access token in Profile → API Keys and update your client config.
404 Tool not found / agent doesn't know any PulseGuard tools
The client has not completed the MCP initialize handshake, or has a stale tool cache.
- Claude Desktop / Windsurf: fully quit and relaunch the app.
- Cursor: open the MCP settings panel and click Reconnect.
- Verify the config file has no JSON syntax errors (trailing commas are not allowed).
[ ] Empty monitor list or "Monitor not found"
The tool only returns monitors owned by the token's user.
- Log into PulseGuard and confirm monitors exist on the dashboard.
- If passing a
monitor_id, copy it directly from the URL orlist-monitors-tooloutput — IDs are UUIDs. - No ping data for stats: the monitor must have at least one completed check cycle.
ERR Connection refused / server not responding
Check the URL in your config:
https://pulseguard.saniolabs.com/mcp
- No trailing slash.
- No
/api/prefix — the endpoint is at/mcpdirectly. - Run the
curlcommand from the Generic HTTP tab above to confirm connectivity from your machine.
Changelog
2026-04-02
Unified Navigation & New Tools
- Unified marketing navigation across all pages (x-marketing-nav)
- Added
list-all-incidents-toolfor bulk incident reporting - Monitor deletion is now available only with Full Access, an eligible paid plan, and the exact monitor ID
- App logo now links to homepage; simplified auth-aware CTAs
- Removed forced dashboard redirect for logged-in users
2026-03-30
Initial release
- SSE-based Remote MCP server at
/api/mcp/sse - 5 monitors & stats tools
MCP server version: 1.61.0 · Protocol: MCP 2024-11-05 · Transport: SSE