Smart Columns

A smart column is a column that executes work per row — instead of just holding data. Hypertab ships eight kinds: static, AI, HTTP, formula, integration, waterfall, lookup, and extract.

Each kind has its own configuration form, rate-limit profile, and retry policy. AI agents create them via MCP; humans create them through the column-add dropdown.

The seven smart column kinds

(Plus one — static — that just holds data.)

AI columns

aggressive (500 req/s) by default

Per-row AI prompt with templated context. Bring your own OpenAI / Anthropic / Google key.

// example
Score each lead 0-100 based on website copy: {{website}}

HTTP columns

moderate (50 req/s)

Call any REST/GraphQL endpoint per row. Extract specific JSON paths into typed cells.

// example
GET https://api.clearbit.com/v2/companies/find?domain={{domain}}

Formula columns

no limit (local)

Compute from other columns using a JS-like expression engine. Free tier — never counts as ops.

// example
Math.round({{revenue}} * 0.15)

Integration columns

conservative (10 req/s)

Push to HubSpot, Salesforce, Slack, etc. when conditions are met.

// example
When {{score}} > 80, push row to HubSpot deal pipeline

Waterfall columns

gentle (2 req/s)

Try sources in order, use the first one that returns data.

// example
Try Apollo → Hunter → ZoomInfo for email enrichment

Lookup columns

no limit (local)

Cross-table VLOOKUP with optional caching. Free tier — never counts as ops.

// example
Match deal.account_id to accounts.id, return accounts.name

Extract columns

no limit (local)

Pull a JSON field from an upstream smart column. Zero API cost — just data routing.

// example
From {{enrich_response}} extract .company.industry

Reliability built in

Token-bucket rate limiting — proactive throttle per (provider, account).
Adaptive backoff on 429 — auto-reduce rate when APIs throttle.
Circuit breaker — pauses column if >50% error rate in 60s window.
Retry with exponential backoff — configurable max retries per cell.
Per-cell state tracking — idle → pending → running → complete/error/skipped.
Cross-row parallelism — plan-gated 5 → 200 concurrent row DAGs.

FAQ

What's a smart column in Hypertab?: A smart column is a column that executes work per row instead of just holding data. There are eight kinds: static, AI, HTTP, formula, integration, waterfall, lookup, and extract. Each has its own configuration and rate-limit profile.
How do smart columns reference each other?: Use template syntax: {{column_name}} or {{column_name.field.path}} for JSON paths. Hypertab parses these refs and auto-builds a dependency graph (DAG). You can also explicitly declare dependencies via config.depends_on.
What happens on errors?: Each cell tracks state: idle → pending → running → complete/error/skipped. Errors don't stop the run — downstream cells are marked skipped with a reason. You can retry failed cells via hypertab_retry_failed_cells. Circuit breaker pauses the column if >50% error rate in 60s.
How is rate limiting handled?: Token-bucket per (provider, account). Adaptive throttling auto-reduces rate on 429 responses. Presets: aggressive (500/s), moderate (50/s), conservative (10/s), gentle (2/s). You can override per-column.

Build your first smart column

Free tier covers 5K ops/month. No credit card.

get started free →