@tablebi/cli

@tablebi/cli

The TableBI command line — the integration surface for your Claude Code / Codex. TableBI connects scattered marketing/SEO sources, normalizes them into one trusted set of numbers (口径), and lets your agent query, build, and publish live dashboards.

npm i -g @tablebi/cli
tablebi login              # browser auth → caches your account's default workspace
tablebi install            # plant the TableBI skill into Claude Code / Codex
tablebi context --json     # rehydrate: 口径 + sources + freshness + command cheatsheet

tablebi context needs no -w: logged in, it targets your account's primary workspace (local, it's default). See every workspace you can use — and the default — with tablebi workspaces --json; tablebi whoami --json adds your user + org. Pass -w <id> to target a specific one. (No login? Every command runs against the local .tablebi/ store.)

What it does — Pipe → Ask → Pin

• Pipe — tablebi connect <provider> · discover · sync: OAuth in sources and pull them into raw-complete facts (Google Ads, Meta, TikTok Ads, GA4, Search Console, CSV…).
• Ask — two query altitudes over the same data:
  • tablebi metrics --metric roas --group-by campaign --json — unified 口径 (cross-channel).
  • tablebi query --platform tiktok_ads --select video_views --group-by ad --json — native, full-granularity, single-source (nothing the platform returned is lost).
  • Explore first: tablebi schema · values · sample.
• Pin — tablebi dashboard publish → a hosted, read-only URL that always shows the latest data. Author specs with computed metrics (computed), preview them live (dashboard preview), and embed native fields (source: native).

Built to be driven by an agent: every read takes --json, and the bundled skill teaches your Claude Code / Codex when and how to use each command. There is no LLM in this package — the analysis brain stays your own agent; TableBI provides the trustworthy data + semantic layer.

tablebi --help lists every command.