Datadog Observability — Gram Project

Gram Services

Always filter by the relevant service(s) when querying Datadog:

Available Tools

Use only the following Datadog MCP tools unless the user explicitly asks for others:

Logs

• mcpdatadog-mcpsearchdatadoglogs — Search and filter log events
• mcpdatadog-mcpanalyzedatadoglogs — Analyze log patterns and aggregate stats

Metrics

• mcpdatadog-mcpgetdatadogmetric — Get a specific metric's time-series data
• mcpdatadog-mcpgetdatadogmetric_context — Get context and metadata for a metric
• mcpdatadog-mcpsearchdatadogmetrics — Search available metrics by name

Traces & Spans

• mcpdatadog-mcpgetdatadogtrace — Get a specific trace by ID
• mcpdatadog-mcpsearchdatadogspans — Search spans (useful for latency investigation)

Incidents & Monitors

• mcpdatadog-mcpsearchdatadogincidents — Search active/recent incidents
• mcpdatadog-mcpgetdatadogincident — Get details for a specific incident
• mcpdatadog-mcpsearchdatadogmonitors — Find monitors and their current state

RUM & Events

• mcpdatadog-mcpsearchdatadogrum_events — Search Real User Monitoring events (frontend errors, sessions)
• mcpdatadog-mcpsearchdatadogevents — Search Datadog events stream

Services & Infrastructure

• mcpdatadog-mcpsearchdatadogservices — Discover services in APM
• mcpdatadog-mcpsearchdatadogservice_dependencies — View service dependency map

Guidelines

• Always scope queries to one or more Gram services using the service filter when available.
• Start narrow, expand if needed: Query a 15–30 minute window first, then widen.
• For error investigations: start with searchdatadoglogs, filter by status:error, then follow trace IDs with getdatadogtrace.
• For latency issues: use searchdatadogspans with service:<name> and sort by duration.
• For frontend issues: prefer searchdatadogrum_events for gram.
• For incidents: check searchdatadogincidents first before deep-diving into logs.