Your terminal, your observability data.
The PandaProbe CLI is a single binary that gives you full access to the PandaProbe API from your terminal. Inspect traces, trigger evaluations, and manage sessions without opening a browser.
- List and inspect traces, sessions, and spans
- Create evaluation runs and submit scores
- Filter and sort observability data server-side
- JSON and table output formats — pipe to jq
- Shell completion: bash, zsh, fish, PowerShell
- Auth: browser login for Cloud; API key for self-hosted
- Standardized exit codes for scripting and automation
Everything you need, at the prompt.
Supports pagination, server-side filtering, and up to 200 results per request. Results stream as JSON or human-readable tables.
List, get, and filter traces by status, user, session, or tags. Drill down to span-level detail by kind.
Manage conversation and workflow records. Include full trace data or filter by session metadata.
Read available metrics, create trace or session eval runs, submit scores, and batch-run evaluations.
Generate shell completion scripts for bash, zsh, fish, or PowerShell to speed up your workflow.
The PandaProbe Skill
The PandaProbe CLI ships with a pre-built skill pack that exposes the full CLI as a tool any coding agent can call. Your Claude Code, Cursor, or VS Code agent can list traces, trigger evaluations, and read scores — without ever leaving the IDE.
The CLI was designed from day one with agents as the primary interface. Every command produces structured JSON output that agents can parse, reason over, and act on in subsequent steps.
Why agent engineers reach for the skill
- Agent-native: designed to be called by coding agents, not just humans
- Structured output: JSON responses agents can parse and act on
- Composable: chain CLI commands in agent tool sequences
- Any coding agent: works with Claude Code, Cursor, VS Code agents, and more
- Trigger eval runs, inspect traces, and read scores — all from inside the IDE