Supaboard vs TextQL

TextQL pitches itself as the "sovereign ontology" behind an AI analyst called Ana: a chat-first agent that sits on top of a git-stored semantic layer which is supposed to get cheaper and more accurate every time it's used. Unlike most enterprise AI-analyst vendors, TextQL does publish tiered pricing on its own site: a free Analyst tier plus usage, a $250-plus-usage Team tier, and a custom Enterprise tier, all metered by compute rather than seats. The real question isn't whether the ontology story is credible (TextQL's own "35.7% more accurate" figure on its homepage measures the same agent's accuracy gain from reusing the ontology versus starting from scratch, not an absolute accuracy score) — it's whether a compute-metered bill and a two-surface chat presence cover what the rest of the business needs.

Where the cost actually lands

TextQL's Team tier includes $250/month in credits at $0.003 per ACU (about 83,000 ACUs), with overages at $2-3 per 1,000 ACU depending on the plan. That means the bill tracks how much the org actually queries — a heavy month costs more, a light one costs less — and only Enterprise gets a flat negotiated number, reached through a demo request. Supaboard's $99 (or $83 billed annually) per seat per month bakes usage credits into the seat price, so a finance team isn't forecasting ACU consumption to budget for it.

What is actually different

TextQL's Ana breaks large or multi-source questions into parallel subagents that sweep, filter, and aggregate at scale — real multi-step execution. What TextQL doesn't publish is a named "why did this move" or forecasting mode; Supaboard's Deep Dive is built specifically to answer causal and forecast questions, not just decompose big queries into parallel workers.

Data apps are genuinely prompt-built at TextQL and stored back into the ontology. Dashboards aren't a separately named product, though — TextQL's own site groups "data apps and dashboards" into one category, so it's unclear whether a standalone dashboard is a first-class one-prompt object the way it is in Supaboard, which generates a full live dashboard from a single prompt.

TextQL's Playbooks/agents pattern can detect issues (its own published example: stale pricing rows) and deliver a brief on schedule to Slack or Teams as an Adaptive Card. Alerts are listed among what Ana can deliver, but there's no documented threshold or anomaly configuration screen — it reads more like an agent noticing something mid-run than a standing plain-English rule, which is how Supaboard describes its alerts.

Both handle cross-source questions: TextQL explicitly advertises joining warehouse history with live SaaS data in one question, the same idea behind Supaboard spanning every connected source. On Python, TextQL's own docs are explicit that "you never need to write code yourself" — the agent runs Python internally for charts and analysis, with no user-facing editor. Supaboard puts SQL and Python in the same editor for people who want to write and run it themselves, next to the natural-language layer.

TextQL's MCP story is genuinely strong and verifiable both ways: it exposes an MCP server (with ana, ana_ask, and list_connectors tools) and consumes external MCP servers like Notion or GitHub. What's not documented is a native presence inside Claude, ChatGPT, or Cursor — TextQL ships Slack and Microsoft Teams, while Supaboard lists five shipped surfaces (Slack, Teams, Claude, ChatGPT, Cursor).

On breadth, TextQL publishes 63 connectors (18 SQL, 45 API). Supaboard's catalogue is over an order of magnitude larger (700+, 783 total in the catalogue, 124 connectable without contacting Supaboard) — a material difference if the org's data lives across a long tail of niche SaaS tools rather than a core warehouse. Neither ships an interactive SQL/Python editor with schema-aware autocomplete for end users on TextQL's side; Supaboard's editor lets an analyst review AI-proposed SQL/Python edits as a diff before approving them.

What to test if you trial both

Feed both tools the same multi-source question — one that needs a warehouse plus a SaaS connector neither vendor has seen before — and judge not just whether the query is correct, but whether the tool hands back a usable chart, alert, or scheduled report, or just an answer someone still has to package. Then run a real week of usage against TextQL's ACU-metered bill against Supaboard's flat seat price with credits included, since a short pilot rarely shows how compute-based pricing behaves at production query volume, and TextQL's own pricing page routes anything past the Team tier to a demo request rather than a self-serve checkout.