AxioRankDocs
Integrations

Datadog

Stream the AxioRank governed audit log into Datadog and install the AxioRank tile for a prebuilt log pipeline, two dashboards, and four monitors. Watch agent governance next to the rest of your stack.

Most AxioRank integrations govern the calls your agents make. The Datadog integration is on the observability side: it ships the governed audit log into Datadog so every agent tool call sits next to the rest of your telemetry. It has two halves, configured independently:

  1. The log stream you turn on in AxioRank, which pushes events to the Datadog Logs intake.
  2. The AxioRank tile in Datadog, which adds a log pipeline, two dashboards, and four recommended monitors over those events.

An outbound SIEM destination, not an inbound check

Unlike inbound surfaces, which verify the agents reaching into a site you operate, the Datadog integration forwards your governance evidence outward. Nothing in Datadog changes how AxioRank decides a call. It is read-only telemetry.

Turn on the log stream

  1. In Datadog, copy an API key from Organization Settings, API Keys.
  2. In AxioRank, open Settings, Log Streaming, add a Datadog destination, paste the key, and set your site (for example datadoghq.com, datadoghq.eu, or us5.datadoghq.com).
  3. Save and click Send test to emit a sample event, then confirm it in Datadog under Logs, Explorer with source:axiorank.

Every event arrives as one structured log. The secrets in each call are already redacted at write time, so your SIEM gets the evidence without the credential.

Install the tile

Find AxioRank on the Datadog Integrations page and click Install. The tile provisions:

  • A log pipeline bound to source:axiorank that maps the AxioRank decision to the Datadog status (deny to error, hold to warn, allow to info), exposes risk_score as a measure, sets the event time from created_at, and maps trace_id to the reserved trace id.
  • AxioRank, AI Governance Overview: decisions over time, deny rate, top blocked agents and tools, risk over time, and shadow-AI discoveries.
  • AxioRank, Threats and Compliance: high-risk events, kill chains by trace, information-flow (taint) blocks, the enforcing policies, egress hosts, and an exportable evidence stream of denied and held calls.
  • Four recommended monitors: deny-rate spike, high-risk kill chain, taint-block surge, and shadow-AI discovery surge.

Facets you get

FacetMeaning
statusDerived from decision: deny to error, hold to warn, allow to info.
@decisionallow, hold, or deny.
@risk_scoreRisk of the call, 0 to 100 (a measure).
@agent_idThe agent that made the call.
@tool_nameThe tool or function invoked.
@matched_policy_idThe policy that decided the call.
@taint_blocked · @taint_tagsInformation-flow blocks and the data provenance behind them.
@trace_idStitches the steps of one agent run into a kill chain.
@event_type · @providerShadow-AI discovery findings and the provider involved.

What the dashboards do and do not show

Spend and cost are not part of the streamed audit log today, so there is no cost widget. The "inbound" style signals on the dashboards come from information-flow blocks (@taint_blocked) and streamed shadow-AI discovery findings, not a separate inbound feed. Cost and spend live in the AxioRank console, not in this stream.

Next steps

  • Audit export: the same log as NDJSON or CSV, on demand.
  • Audit integrity: what each event records and how it is signed.
  • Policies: the rules that produce the decisions you see in Datadog.

On this page