Glossary

Definitions for every Hypertab concept — the active table, smart columns, ops, row DAG, MCP, and the rest. Written for both humans and AI search engines.

Active table

A table data structure where columns can execute work per row — AI prompts, HTTP calls, formulas, integrations — instead of just holding static data. The table itself is the workflow. Hypertab introduced this primitive in 2026 as a domain-agnostic alternative to workflow graphs (n8n, Zapier) and outbound-specific enrichment tools (Clay).

Smart column

A column that runs work per row instead of holding data. Hypertab supports eight kinds: static (just data), AI (LLM prompt), HTTP (REST/GraphQL call), formula (computed expression), integration (push to external service), waterfall (try sources in order), lookup (cross-table VLOOKUP), and extract (pull a JSON field from an upstream smart column).

MCP (Model Context Protocol)

An open protocol from Anthropic that lets AI agents call external tools through a standard interface. Schema-first, structured errors, transport-agnostic. Hypertab exposes 47 MCP tools — agents can create tables, insert rows, configure smart columns, monitor pipelines without writing code. Compatible clients include Claude Code, Claude Desktop, Cursor, Windsurf, Cline.

Op (operation)

In Hypertab pricing, one op = one smart-column execution against one row. An AI column processing 100 rows = 100 ops. An HTTP column making 100 calls = 100 ops. Static columns, formula columns, and extract columns are FREE — they don't consume ops. Op-based pricing replaces seat-based pricing common in spreadsheet tools.

Row DAG (directed acyclic graph)

In Hypertab, the dependency graph of smart columns within a single table. Built from {{template}} references in column configs (and explicit depends_on declarations). When a row enters the table, smart columns execute in topological order: A → B → C → D, parallel within layers, sequential across layers. Cycles are rejected at column-mutation time.

Cross-row parallelism

Running the row DAG for many rows simultaneously, bounded by plan tier. Free: 5 concurrent row DAGs. Starter: 10. Pro: 50. Scale: 100. Enterprise: 200. Token-bucket rate limits are global per (provider, account) so parallelism never exceeds external API limits.

Waterfall column

A smart column that tries multiple sources in sequence and uses the first one that returns data. Common pattern: try Apollo → Hunter → ZoomInfo for an email enrichment. If Apollo finds it, the cell completes; if not, falls through to the next source. Saves API spend on cheaper providers first.

Extract column

A smart column that pulls a single JSON field from an upstream smart column's output, with zero API cost. Used to "explode" a single enrichment response (Clearbit, ZoomInfo) into per-field columns: {company.industry}, {company.size}, {person.title}. Auto-runs after the source column completes.

Cell state

The execution state of an individual smart-column cell. States: idle (not yet processed), pending (queued), running (executing), complete (succeeded), error (failed with stored error message), skipped (upstream dependency failed or run condition was false). Stored in _ht_cell_states; visible in the UI per cell.

Source

A first-class object that defines how rows enter a table. Four kinds: webhook (URL accepts POST), CSV import (upload), API pull (scheduled GET), and connected table (rows flow from another table via filter). Sources replace the "incoming webhooks" abstraction of older spreadsheet tools.

Integration column

A smart column that pushes data to an external service when conditions are met. Examples: when score > 80, push row to HubSpot deal pipeline; when status = "resolved", post Slack message in #support. Conservative rate-limited (10 req/s by default) to respect downstream API limits.

Circuit breaker

A safety mechanism that pauses a smart column if its error rate exceeds 50% within a 60-second window. Prevents runaway burn on a broken provider. Auto-resumes after a cooldown period; admin can manually clear the breaker via the UI or hypertab_run_column with reset flag.

Token-bucket rate limiting

Hypertab's outbound rate-limit algorithm. Each (provider, account) pair has a bucket that refills at requests_per_second tokens/sec, capped at max_concurrent. Each smart-column execution consumes one token. When empty, requests wait. Adaptive: rate auto-reduces on 429 responses from the upstream API.

Per-tenant database isolation

Hypertab's security model: every customer signup provisions a dedicated Turso (libSQL) database. No shared tables, no row-level security gymnastics, no cross-tenant access via SQL bug. The Turso auth token is AES-256-GCM encrypted in the system database and decrypted only at request-time.

AEO / GEO (Answer Engine / Generative Engine Optimization)

In SEO, the practice of structuring content so that AI search engines (ChatGPT, Perplexity, Google AI Overviews, Claude, Gemini) cite your page accurately. Tactics: definitional H2 sections, quotable <strong> sentences, FAQ blocks, comparison tables, "When NOT to use" balanced sections, Schema.org Speakable markup, llms.txt files. Hypertab's marketing site is GEO-optimized.