generated: '2026-09-23'
method: searched
source: https://docs.datadoghq.com/cli.md
name: Pup CLI
description: >-
  Pup is Datadog's official AI-agent-ready command-line interface. It exposes
  the Datadog API surface for use in terminal and AI-agent workflows, structures
  commands so agents can navigate them without external documentation, supports
  OAuth2/PKCE for scoped authentication (falling back to API keys), and returns
  JSON/YAML for machine consumption.
repository: https://github.com/DataDog/pup
install:
  - method: homebrew
    platforms: [macOS, Linux]
    commands:
      - brew tap datadog-labs/pack
      - brew install datadog-labs/pack/pup
  - method: source
    commands:
      - git clone https://github.com/DataDog/pup.git && cd pup
      - cargo build --release
      - cp target/release/pup /usr/local/bin/pup
  - method: binary
    url: https://github.com/DataDog/pup/releases/latest
authentication:
  primary: OAuth2 (browser flow via `pup auth login`), with PKCE
  fallback: API keys via DD_API_KEY / DD_APP_KEY
  stateless: DD_ACCESS_TOKEN bearer for WASM/headless
commands:
  - group: auth
    examples:
      - pup auth login
      - pup auth login --org
  - group: monitors
    examples:
      - pup monitors list --tags="team:api-platform"
  - group: logs
    examples:
      - pup logs search --query="status:error" --from="1h"
  - group: metrics
    examples:
      - pup metrics query --query="avg:system.cpu.user{*}" --from="1h"
  - group: dashboards
    examples:
      - pup dashboards get <DASHBOARD_ID>
      - pup dashboards delete <DASHBOARD_ID> --yes
product_coverage:
  - Core Observability (Metrics, logs, events, RUM, APM, traces)
  - Monitoring and Alerting (Monitors, dashboards, SLOs, synthetics, downtimes, workflows)
  - Security and Compliance (Security rules, signals, findings, audit logs, CSM threats)
  - Infrastructure and Cloud (Hosts, tags, containers, network, AWS/GCP/Azure integrations)
  - Incident and Operations (Incidents, on-call, work management, error tracking, service catalog)
  - CI/CD and Development (CI visibility, test optimization, DORA metrics, deployment gates)
  - Organization and Access (Users, API keys, application keys, organizations)
  - Platform and Configuration (Usage metering, cost management, feature flags, observability pipelines)
global_flags:
  - flag: -o, --output
    description: 'Output format (json, table, yaml). Default: json.'
  - flag: -y, --yes
    description: Skip confirmation prompts for destructive operations.
  - flag: --agent
    description: Enable agent mode.
  - flag: --no-agent
    description: Disable agent mode.
  - flag: --read-only
    description: Block all write operations (create, update, delete).
  - flag: --org <org>
    description: Use a named org profile for multi-account workflows.
environment_variables:
  - name: DD_ACCESS_TOKEN
    description: Bearer token for stateless authentication.
  - name: DD_API_KEY
    description: Datadog API key (optional if using OAuth2 or DD_ACCESS_TOKEN).
  - name: DD_APP_KEY
    description: Datadog application key (optional if using OAuth2 or DD_ACCESS_TOKEN).
  - name: DD_SITE
    description: 'Datadog site (default: datadoghq.com).'
  - name: DD_AUTO_APPROVE
    description: Auto-approve destructive operations (true/false).
  - name: DD_TOKEN_STORAGE
    description: 'Token storage backend (keychain or file, default: auto-detect).'
agent_mode:
  description: >-
    When Pup is invoked by a supported AI coding agent, it auto-detects and
    switches to agent mode: structured JSON responses, metadata, error details,
    hints, and auto-approval of confirmation prompts. Explicit enable via
    --agent or FORCE_AGENT_MODE=1.
additional_features:
  - runbooks (local execution engine for YAML-defined operational procedures)
  - agent skills (ships domain agents/skills embedded in the binary; installable with `pup skills install`)
  - ACP server (`pup acp serve` runs a local AI agent server; ACP and OpenAI-compatible protocols)
notes: >-
  There is also a legacy Python-shipped `dog` CLI bundled with the `datadogpy`
  library (pypi: datadog); Pup is the current AI-agent-ready CLI documented on
  the docs site.