name	observability-clis
description	Query PostHog error tracking and Sentry issues from the command line. Use when asked what errors are happening, which errors are being missed or untriaged, for production error rates or impact by user, or whenever posthog-cli or sentry-cli is mentioned.

Observability CLIs

Both tools have quirks that waste turns if rediscovered. Org and project slugs are project-specific — check that project's CLAUDE.local.md.

PostHog

The binary is posthog-cli, installed via mise. The agent-first surface is posthog-cli api:

posthog-cli api search <regex>              # find tools
posthog-cli api info --json <tool>          # get a schema
posthog-cli api call --json <tool> '<json>' # run one

For "what errors are happening" or "what is being missed", use query-error-tracking-issues-list. It defaults to active issues, the last 7 days, sorted by occurrences, with test accounts filtered out.

For impact rather than volume:

posthog-cli api call --json query-error-tracking-issues-list \
  '{"dateRange":{"date_from":"-30d"},"orderBy":"users"}'

assignee: null across the board means nobody is triaging.

Sentry

sentry-cli uses a personal read token in ~/.sentryclirc.

The --org flag goes after the subcommand, not before:

sentry-cli projects --org <org> list

There is no issues-list subcommand. Query issues through the REST API with the same token, read from the file inline so the literal stays out of shell history and transcripts:

curl -H "Authorization: Bearer $(sed -n 's/^token=//p' ~/.sentryclirc)" \
  "https://sentry.io/api/0/projects/<org>/<project>/issues/?statsPeriod=24h&query=is:unresolved&sort=freq&limit=15"

Useful query filters: is:unresolved, an endpoint path, N+1.

Parse responses with json.loads(..., strict=False). Bodies contain raw control characters inside SQL strings and strict parsing fails on them.

observability-clis