Datadog Labs

Runs a live check on Datadog instrumentation across your Linux hosts and generates a report with clickable links to the exact metrics in your dashboard.

Writes APM remapping rules to collapse auto-generated or renamed services into a single canonical name, reducing noise in dashboards and alerts without touching code.

Analyzes logs and config on a Linux host running Single Step Instrumentation to pinpoint why traces aren't flowing to your APM tool — covers agent setup, library injection, and permissions.

Runs diagnostics on a Linux host to confirm Single Step Instrumentation is live and metrics are flowing to your APM backend without manual code edits.

Pipes Claude agent logs, latency, errors, and token usage into Datadog dashboards. Surfaces when an agent is degrading before it breaks production.

Walks through agent install, library instrumentation, and service configuration for your language and framework. Outputs setup code and explains what you're measuring.

Queries your Datadog Audit Trail for evidence of SOC 2 and PCI DSS controls, maps findings to framework requirements, and produces a formatted report ready to hand an auditor.

Pulls Datadog audit logs and activity, builds a timeline of suspicious calls, flags geographic and endpoint anomalies, and recommends rotation or revocation steps.

Parses your audit trail to answer "who deleted what" or "when did this config change" — flags unusual IP login patterns, permission escalations, and suspicious bulk actions.

Searches Datadog's official docs and returns exact code snippets, configuration examples, and API details so you don't context-switch to browser tabs.

Generates a ready-to-run Python script or Jupyter notebook pre-wired to DataDog's LLMObs SDK, following their reference style. Just add your prompt and data.

Reads PR logs and attributes each failure to a flake, infrastructure issue, or code regression. Surfaces coverage gaps and flags security or quality concerns in one view.

Installs the Datadog Agent on Linux hosts via SSH and enables automatic application tracing — no code changes needed. Skips hosts that already have it.

Configures Datadog's browser SDK for RUM, logs, session replay, and error tracking. Handles version upgrades, troubleshoots setup issues, and migrates from older versions.

Takes experiment IDs and outputs win/loss rates, statistical significance, failure modes, and the 2–3 changes most likely to have moved the needle.

Configures Single Step Instrumentation to automatically collect performance metrics from your apps without touching code. Requires Datadog Agent already running.

Queries your log archives, surfaces high-volume low-value data streams, and recommends which logs to sample, filter, or drop to cut spend without losing signal.

Logs every tool call the Bits AI assistant made in your Datadog org—who ran it, what resources it accessed, and flags unusual patterns for security review.

Takes your LLM app's untagged traces and runs them through classification, root-cause analysis, and eval generation—producing a ready-to-run evaluator suite with human checkpoints built in.

Reads your recent API logs and generates evaluator code that grades outputs against real patterns from your data. Emits ready-to-run SDK code or JSON specs.

Finds deprecated v6 options in your codebase, removes them, updates CDN references, and flags breaking changes so your monitoring keeps working after the upgrade.

Walks you through the ownership preference types, generates a ready-to-upload CSV, and shows you exactly how to load it — whether via UI, API, Terraform, or cloud storage.

Reads Datadog LLM Obs traces and RUM sessions, flags interactions where the user's intent went unmet, and surfaces patterns across your whole product or a single conversation.

Creates and manages Datadog monitors from plain English, searches your existing monitors, and flags alert rules that are too noisy or miss failure modes.

Reads production trace data—eval failures, errors, or behavioral anomalies—and walks the span tree backward to pinpoint the exact model, prompt, or data step that broke.

Pulls test history and failure patterns, categorizes the root cause (timing, environment, flawed logic), and recommends whether to fix, quarantine, or escalate to ops.

Parses audit logs to surface key events: unauthorized access, cost spikes, config changes, or AI model decisions. Outputs a timeline with enough detail for compliance reports.

Pulls your Datadog Usage Metering and Audit Trail logs, maps the spike in costs to the specific config change that triggered it, and names the person and timestamp.

Connects Claude to your Datadog account via OAuth2, letting you pull metrics, logs, and dashboards without leaving the chat. Token refresh is automatic.