Tested in public

686 of 686 tests pass.

Every feature ships with a record of how it was tested: unit and integration tests in the real Workers runtime, benchmarks on 50,000 rows, and scenarios clicked through in a browser.

FeatureTestsFilesScenariosChecked
Dashboards61 / 614122026-10-21
Calendar, timeline, gallery and grid grouping41 / 41662026-10-21
MCP server35 / 35242026-10-21
Core engine and API272 / 272602026-10-21
Automations125 / 12514122026-10-21
AI fields72 / 7211122026-10-21
Plan limits, account emails and hosting extensions46 / 46472026-10-21
Email code sign-in and sign-in options34 / 34452026-10-21

Dashboards: benchmarks

ActionTimeRows
Dashboard with 8 widgets, cache empty1,401 ms50,000
The same dashboard, cached0.02 ms50,000

local wrangler dev, Apple M5 Pro, node v24.18.1. Local measurements, not production.

Calendar, timeline, gallery and grid grouping: benchmarks

ActionTimeRows
Calendar month window (6 weeks, 2,000 rows inline)75.4 ms50,000
Grid grouped by 2 fields + sum summaries579.33 ms50,000
Kanban group counts64.95 ms50,000

local wrangler dev, Apple M5 Pro, node v24.18.1. Local measurements, not production.

MCP server: benchmarks

ActionTimeRows
Group by a select field with count and sum (aggregate)205.63 ms50,000

local wrangler dev, Apple M5 Pro, node v24.18.1. Local measurements, not production.

Core engine and API: benchmarks

ActionTimeRows
Open view first page (300 ids + count)0.5 ms50,000
Load full id list of a 50k view15.06 ms50,000
Edit a cell + rollup update0.28 ms50,000
Filter by any field53.33 ms50,000
Sort by any field98.4 ms50,000
Filter + sort65.76 ms50,000
Search70.48 ms50,000
Full rollup recompute294 ms50,000
Full formula column recompute251 ms50,000
One 500-row recompute batch2.72 ms50,000

local wrangler dev, Apple M5 Pro, node v24.18.1. Local measurements, not production.

Automations: benchmarks

ActionTimeRows
Edit a cell + rollup update0.2 ms50,000
Edit a cell + rollup update, 3 record triggers on the table0.24 ms50,000
Paste 1,000 rows and delete them24.44 ms50,000
Paste 1,000 rows and delete them, 3 record triggers (1,000 runs recorded)57.17 ms50,000
Execute 100 queued runs that read and update their record31.08 ms50,000
Record edit to outgoing HTTP request, median of 58 ms30

local wrangler dev, Apple M5 Pro, node v24.18.1. Local measurements, not production.

AI fields: benchmarks

ActionTimeRows
Edit a cell + rollup update1.2 ms50,000
Edit a cell + rollup update, 2 auto-updating AI fields on the table0.8 ms50,000
Paste 1,000 rows and delete them132.37 ms50,000
Paste 1,000 rows and delete them, 2 AI fields (2,000 cells queued)121.25 ms50,000
Generate a column of 1,003 records, fake model answering in 1 s, at most 6 calls at once171 s1,003
Record edit to AI answer written back with DeepSeek deepseek-flash, median of 5 (includes 2 s debounce)3,216 ms30

local wrangler dev, Apple M5 Pro, node v24.18.1, machine under other load (load average about 31); local wrangler dev, Apple M5 Pro. Local measurements, not production.