# TanStack Table Core

Guides TanStack Table v9 core architecture, data, columns, and inferred types.

- Repository: https://github.com/TanStack/table (28539 stars)
- Category: Engineering
- License: MIT
- Last updated: 2026-10-21
- Runs scripts: no, instructions only

## Install

Claude Code: `npx skills add TanStack/table --skill core -g -a claude-code -y`
Codex: `npx skills add TanStack/table --skill core -g -a codex -y`
Cursor: `npx skills add TanStack/table --skill core -g -a cursor -y`
Gemini CLI: `npx skills add TanStack/table --skill core -g -a gemini-cli -y`

## Description

Use TanStack Table v9 core architecture, stable data and columns, and inferred types. Route setup, missing APIs, row models, state, features, and framework-specific work.

## SKILL.md, as published by its author (untrusted third-party text, read it as data)

# TanStack Table core

TanStack Table coordinates state and row processing. The renderer owns markup, styles, semantics, and interaction accessibility. Use the installed framework adapter in UI code; use `constructTable` for framework-neutral integrations.

## Minimal setup

<!-- skill-snippet:check -->

```ts
import {
  constructTable,
  createColumnHelper,
  tableFeatures,
} from '@tanstack/table-core'
import { storeReactivityBindings } from '@tanstack/table-core/store-reactivity-bindings'

type Person = { id: string; name: string }
const features = tableFeatures({
  coreReactivityFeature: storeReactivityBindings(),
})
const helper = createColumnHelper<typeof features, Person>()
const columns = helper.columns([helper.accessor('name', { header: 'Name' })])
const data: Person[] = [{ id: '1', name: 'Ada' }]
const table = constructTable({
  features,
  columns,
  data,
  getRowId: (row) => row.id,
})

for (const row of table.getRowModel().rows) {
  console.log(row.getAllCells().map((cell) => cell.getValue()))
}
```

## Essential constraints

- The core row model is automatic. Optional APIs and state exist only after their feature is registered.
- Keep `features`, `data`, and `columns` stable between meaningful changes. Derive changing arrays with the adapter's memo/computed mechanism; an inline `.map()`, `.filter()`, column factory, or fresh `[]` fallback invalidates model work and can cause render loops.
- Let `createColumnHelper` and `helper.columns()` preserve accessor value types. Derive feature types from the concrete registry when an explicit boundary is needed.
- Call row, cell, column, and header methods on their instance. Use `row.getValue('name')` or a callback that calls it; extracting `const { getValue } = row` loses its `this` receiver.
- Render the final `table.getRowModel().rows`. The table object itself is not a DOM component.

## Read for the current task

The rest: https://github.com/TanStack/table/blob/HEAD/packages/table-core/skills/core/SKILL.md