Creating Rules
Define alert rules with metrics, operators, thresholds, and time windows.
Creating Rules
Creating an alert rule
- Go to Alerts in the sidebar
- Click New Rule
- Fill in the rule fields:
- Name — human-readable label for the rule
- Server — the MCP server to monitor (or "All servers")
- Metric — what to measure
- Operator —
>,>=,<,<=,== - Threshold — the value to compare against
- Window — the time window to evaluate over (minutes)
- Optionally configure a notification channel
- Save
Available metrics
| Metric | Description |
|---|---|
error_rate | Percentage of tool calls with status = error (0–100) |
avg_latency_ms | Average tool call duration in milliseconds |
p95_latency_ms | 95th percentile latency in milliseconds |
call_count | Total tool calls in the window |
server_down | Boolean — 1 if server status is down, 0 otherwise |
Example rules
Alert when any server goes down:
Alert when error rate exceeds 5% on a specific server:
Alert when p95 latency exceeds 2 seconds:
Enabling / disabling rules
Toggle the active state on any rule in the list. Disabled rules are not evaluated.