# Find the cause of an alert or incident across metrics, logs and traces

> Find the cause of an alert or incident across metrics, logs and traces: 8 companies fully do it, 4 with a limit, 15 carry the tag. Every listing in Observability & incidents with the reason and a source.

- Canonical: https://www.anchorterminal.com/niches/observability.investigate-alert/
- Markdown: https://www.anchorterminal.com/niches/observability.investigate-alert/index.md (~1,450 tokens)
- Slim: https://www.anchorterminal.com/niches/observability.investigate-alert/index.min.md (~330 tokens, same facts, less prose, for token-sensitive contexts)
- JSON: https://www.anchorterminal.com/niches/observability.investigate-alert/index.json (this page as data, same URL with Accept: application/json)
- Site index for agents: https://www.anchorterminal.com/llms.txt (the key docs and every listing in one file, at most 16 MB, about 4 million tokens: https://www.anchorterminal.com/llms-full.txt)
- API: https://www.anchorterminal.com/api/v1/index.json
- MCP: https://www.anchorterminal.com/mcp (the whole catalogue as tools, no key)
- DNS: dig +short TXT <slug>.grade.anchorterminal.com (the grade as a DNS TXT record, for a sandbox with no HTTP where names still resolve; also top., best.<category>. and <a>.vs.<b>. under the zone, and the same under grade.letme.dev)
- Licence: CC BY 4.0, credit Anchor Terminal and link this page (https://www.anchorterminal.com/terms/#data)
- Updated: 2026-11-15

Four or more companies' docs show a full fit.

| Companies, full fit | With a limit | Carry the tag | Share of the category's work | First-hand accounts |
| --- | --- | --- | --- | --- |
| 8 | 4 | 15 | 26 | 13 |

## What counts as done

A short report naming the likely cause, the queries or evidence behind it (series, log lines or trace ids with timestamps) and the time window, built only from read calls.

Measured, for our own tests: report within 15 minutes and 30 tool calls.

How agents ask for it: why did this alert fire; why is the app slow; debug this incident; what changed before latency went up

## Does the job

The docs show it meets the test above, and nothing we read says otherwise.

- **Axiom MCP Server (Axiom, Inc.)**: queryDataset takes APL and queryMetrics takes MPL directly; getMonitorHistory and services-getChanges give the alert and what changed; query cost limits return a 429 naming the limit. Accounts acc-15 and acc-16 describe agents finding traces and alerts this way.
- **Better Stack (Better Stack, Inc.)**: Execute query runs ClickHouse SQL over logs, spans, metrics and exceptions; incident details and timeline give the alert; Plan query windows splits long ranges; a Read token keeps it to reads. acc-14 uses it daily to resolve reported errors.
- **Causely MCP server (Causely, Inc.)**: get_alerts, investigate_alert, get_issues and get_diagnosis_details return the inferred cause with causal chain, blast radius, logs, exceptions and events, and get_diagnoses takes a start and end window; all reads.
- **Datadog MCP Server (Datadog, Inc.)**: Core toolset searches logs, spans, metrics, monitors and events with Datadog query syntax, analyze_datadog_logs takes SQL, and get_datadog_trace follows a trace. Accounts acc-05, acc-06 and acc-07 describe agents debugging production this way.
- **Grafana MCP Server (Raintank Inc., d/b/a Grafana Labs)**: PromQL, LogQL and TraceQL tools take native queries, alerting_rules_read and OnCall alert groups give the alert, get_tempo_trace follows a trace id, and --disable-write keeps it to reads. acc-08, acc-11 and acc-12 run incident work this way.
- **Honeycomb MCP (Hound Technology, Inc. d/b/a Honeycomb)**: run_query takes a Honeycomb query spec, get_triggers gives the alert, get_trace follows a trace id and run_bubbleup isolates what differs, all read tools. acc-09 and acc-10 describe alert and regression investigations that return a full report.
- **New Relic MCP Server (New Relic, Inc.)**: execute_nrql_query takes NRQL directly; search_incident, list_recent_issues and list_change_events give the alert and what changed; search_errors takes a trace id and get_distributed_trace_details follows it. The server is read-only by scope.
- **Sentry MCP (Functional Software, Inc. d/b/a Sentry)**: search_errors, search_traces and search_logs, get_trace_details and get_sentry_resource with an issue or trace URL give errors, traces and logs with ids; find_metric_monitors gives the alert. acc-05 and acc-16 use it for incidents and inbound alerts.

## Does the job, with a limit

It does the job, but something in the docs or a user's account will get in an agent's way. The limit is named in each row.

- **Cronitor MCP server (Cronitor)**: get_status, list_issues and get_metrics show which job, heartbeat or check failed and its latest event, but there are no logs or traces, so causes outside the monitored job are out of reach.
- **incident.io API (Pineapple Technology Ltd.)**: alert_show, incident_show and investigation_sync give the alert, linked incidents and findings, but telemetry is reached only through ask_telemetry, a natural-language agent that needs the Investigations product, so queries and series are not the agent's own.
- **PagerDuty MCP Server (PagerDuty, Inc.)**: browse_incidents lists alerts, notes, change events and related or past incidents, but PagerDuty holds no metrics, logs or traces, so the cause needs another server.
- **Rootly MCP Server (Rootly Inc.)**: core and alerts toolsets read incidents and alerts, and an observability toolset reads Rootly Observability logs and patterns; there are no metrics or traces, and logs exist only for teams on Rootly Observability.

## What stops agents finishing

- the relevant signal was never collected
- agent misreads which layer an alert belongs to
- query-cost or rate cap reached mid-investigation
- tool list fills the context before the first query

## Around it

Before: notifications.alert-received, messaging.slack-thread, observability.oncall-and-incident-records. After: code.open-pr, observability.oncall-and-incident-records, messaging.post-summary.

## Where this comes from

Weight 26 of 100 in Observability & incidents: 13 of 42 accounts (about 26 per cent of the 50 account-to-use-case links); the job most often described in the Ask HN MCP-in-production thread and the SRE threads

As data: https://www.anchorterminal.com/api/v1/use-cases/observability.investigate-alert.json, or the MCP tool get_use_case with id observability.investigate-alert. Draft, read 2026-11-15 from public docs and first-hand accounts, not yet reviewed by a person.

---

Query this catalogue directly over MCP: https://www.anchorterminal.com/mcp (Streamable HTTP, no key). How: https://www.anchorterminal.com/agents/#mcp
