Pup is Datadog’s official open-source CLI that gives AI agents full access to the Datadog observability platform. Published on GitHub under the Apache-2.0 license with about 1,031 stars and actively developed in 2026, Pup exposes more than 200 commands across 33+ Datadog products in a form agents can navigate reliably. As Datadog puts it, every AI agent needs a loyal companion — this one fetches metrics, sniffs out errors, and tracks down production issues.
Pup is written in Rust and designed from the start for agent-driven infrastructure management. Its commands are self-discoverable, so an agent can explore the available surface without chasing documentation, and output is structured JSON or YAML for easy parsing. Access is secured with OAuth2 and PKCE, giving scoped, leashed permissions instead of long-lived API keys. Through Pup, an agent can work with monitors, logs, metrics, real user monitoring, security signals, and much more. Humans are welcome too — the same CLI works interactively in a terminal for engineers who prefer scriptable access to Datadog.
Datadog built Pup because agent-driven operations are becoming a primary interface to infrastructure: instead of a human clicking through dashboards, an agent investigating an incident can pull the exact metric, log line, or monitor state it needs. Pup’s command surface is deliberately predictable and machine-readable, and its OAuth-based flow means organizations can grant an agent precisely the scopes a task requires — and revoke them just as precisely — which is essential for safe autonomous operations.
Pup itself is free and open source. It operates against a Datadog account, and Datadog is a paid observability platform, so the listing is marked Freemium.
Platform and SRE teams who want their AI agents — or their own scripts — to investigate incidents, query observability data, and manage Datadog resources programmatically.