Priced on the backend.
Not on inference.
You pay for connected sources, two-altitude querying, and pinned live dashboards. You never pay us for AI reasoning — your own Claude Code / Codex is the brain. Every plan: $0 inference, always.
Operator
Your agentFor solo operators and growth engineers driving their own Claude Code on live data.
- → 1 workspace
- → All sources (Google Ads, Meta, TikTok, GA4, Search Console, Sheets, CSV) — connected & auto-synced
- → Two altitudes: unified metrics (trusted cross-channel) and raw (full-granularity, lossless)
- → Pinned live dashboards — hosted read-only URLs that refresh themselves
- →
tablebi install+ CLI / MCP, every command speaks--json - · Bring your own Claude Code / Codex — $0 inference
Agency
Multi-clientFor operators running several brands — one backend, an isolated workspace per client.
- → Up to 10 client workspaces, isolated by row-level security
- → Separate metric definitions, sources & pinned dashboards per client
- → Share any dashboard as a read-only URL — no login for the viewer
- → Everything in Operator, across every workspace
- · White-label client reports & scheduled delivery — coming later
Studio
ScaleFor larger studios running many brands, with priority sources and hands-on onboarding.
- → Up to 40 client workspaces
- → Priority source requests & new connectors
- → White-glove onboarding & metric setup
- → Priority support
- → Everything in Agency
We host the data. You bring the brain.
TableBI hosts no LLM. We give your agent the boring, load-bearing layer — connected sources, one trusted set of numbers, the lossless raw underneath, and a place to pin a live dashboard. The reasoning runs on your own Claude Code / Codex, on your key. So inference is always $0, and every number stays reproducible and traceable — no model in the loop to second-guess.
It's a TypeScript stack. Deploy it on Fly for the price of a coffee.
The engine scales to zero when idle — only a small Postgres stays on. Expected ~$5–15/mo, and LLM cost stays $0 because the brain is still your own agent. Same backend, your infrastructure.
Do I need to know how to code?
+
No. You drive TableBI from your own Claude Code / Codex in plain English — "connect Acme's GA4," "blended ROAS by channel last week," "pin this as a Monday dashboard." tablebi install plants a skill so your agent knows the commands and your metric definitions, then runs them for you. No prompt-engineering, no SQL.
What does $0 inference mean?
+
We host no LLM. TableBI is the data backend — connectors, unified metrics, raw, pinned dashboards — and the reasoning runs on your own agent, on your key. You never pay us for AI, and we never pay for it on your behalf. Every number the engine returns is computed deterministically, so it stays reproducible and traceable.
What's the difference between unified and raw?
+
Two altitudes over the same data. Unified metrics are the cross-channel set of numbers — one definition of cost, conversions and ROAS, reconciled across every platform; the number you quote a client. Raw is single-platform and full-granularity — every native field the platform reports (Video views, Profile visits, anything), kept lossless at ingest. Your agent asks schema to see what each offers, then picks. You never trade trust for granularity.
Can I share a dashboard?
+
Yes. When your agent pins an analysis, TableBI hosts it at a read-only URL — anyone with the link can view it, no login needed. It stores the query, not a snapshot, so the dashboard refreshes itself as new data syncs. Editing happens back in your Claude Code by changing the spec; the page itself stays view-only.
Which sources can I connect?
+
Google Ads, Meta, TikTok, GA4, Google Search Console, Google Sheets, and CSV. GA4, Search Console and Google Ads are live today via OAuth; Sheets and CSV work anytime; Meta and TikTok slot in on the same connector framework — auth, incremental fetch, an encrypted credential vault — so new sources keep arriving. All of them land in one trusted set of numbers, with the raw kept underneath, on every plan.