D

Agent Observability

Developer tool

by Datadog

Traces, evaluates and monitors LLM and AI-agent applications in production, with offline experimentation on datasets built from real traces.

Find alternatives to Agent Observability

Free tier at 40K LLM spans/month with 15-day retention; Pro from $160/month for 100K LLM spans, plus per-span overage and retention add-ons; billed on LLM spans only. Figures come from the product page's own pricing block, not the /pricing/ page, which lists the product without figures. NAME HISTORY, DISCLOSED AS INFERENCE: the page carries no explicit renaming statement; what is verifiable is that the page title reads 'Agent Observability | LLM Observability | Datadog', the H1 is 'AGENT OBSERVABILITY', and https://www.datadoghq.com/product/llm-observability/ serves this page. Reading it as a rename of 'LLM Observability' is this dataset's inference from that evidence, not a vendor statement.

What it does

• Evaluation and observability
• Guardrails and safety

How it compares

• Elastic Agent Builder

  such as AppDynamics (owned by Cisco Systems), Datadog, Dynatrace, New Relic, and Splunk (owned by Cisco Systems).

  Regulatory filing · 18 Sep 2026

Sources

Pricing

Pricing page · 3 Sep 2026

Pricing table published on the product page itself; cross-checked against docs.datadoghq.com/llm_observability/setup/ which confirms span-volume billing but gives no rate. Page and prices independently re-fetched.

Sold within

Pricing page · 29 Sep 2026

Read in the in-app browser 2026-09-29: the rate card prices it on its own row: "Agent Observability For first 100k LLM spans, per month, with 15-day retention by default $160"; the product page offers a free start.

Also from Datadog

• Bits Chat Application

  Conversational AI interface for querying Datadog metrics, logs, traces and monitors in natural language and generating dashboards and notebooks, accessible from Datadog, Slack or mobile.

• Bits Investigation AI agent

  AI SRE agent that autonomously investigates every alert the moment it fires, explores multiple root-cause hypotheses in parallel and reports findings into Slack, Jira, ServiceNow or GitHub.

• Bits Code AI agent

  Coding agent that triages production errors, regressions and vulnerabilities from Datadog telemetry, generates fixes with unit tests grounded in logs, traces and runtime variables, and opens pull requests for review.

• Bits Agent Builder Agent platform

  No-code builder for custom AI agents that investigate, decide and act inside Datadog to automate incident response, observability, security and operational workflows.

• Bits Security Analyst AI agent

  Always-on AI SOC analyst that autonomously triages and investigates security alerts and delivers written investigation results to Datadog, Slack or Jira within minutes.

• Datadog MCP Server Developer tool

  Model Context Protocol server that gives AI coding agents such as Claude Code, Cursor and Codex secure real-time access to Datadog logs, metrics and traces under existing RBAC controls.

• AI Impact Application

  Detects AI-assisted pull requests from coding assistants such as Claude Code, Cursor and GitHub Copilot and compares them on adoption, PR throughput, cycle time, change failure rate and daily cost per active user.

• Watchdog Application

  Datadog's built-in AI engine: it continuously analyses metrics, traces and logs across the platform to raise anomaly alerts without configuration, detect faulty deployments by comparing code versions, run automated root-cause analysis on critical failures and surface tag-based insights and impact analysis. Available inside Infrastructure Monitoring, APM, Log Management and RUM rather than sold on its own.

• GPU Monitoring Application

  Datadog's Infrastructure-family product for shared GPU fleets across cloud, on-prem and neocloud providers: it links device health, cost and performance to the workloads and teams using them, alerts on unmet GPU requests, thermal throttling and ECC/XID errors, forecasts GPU demand and recommends optimisations such as reclaiming GPUs held by zombie processes. The page markets alerting, forecasting and recommendations but does not name a model or AI mechanism behind them.

Open Datadog in the directory

Something wrong here? Send a correction — quote this product id: datadog-agent-observability.