Skip to contents

Getting Started with airtable2

Source: vignettes/airtable2.Rmd
airtable2.Rmd

Overview

airtable2 is an httr2-based R client for the Airtable REST API. It provides:

  • High-level functions (air_*) for reading, writing, upserting, and syncing records with automatic pagination, type coercion, and schema caching.
  • Low-level wrappers (at_*) that map directly to Airtable API endpoints for advanced use.
  • A DBI interface for use with the RStudio/Positron connection pane.
  • Backup and restore functions (air_dump() / air_restore()) for full base exports.
  • Special handling for Airtable-specific types such as multi-select, attachments, collaborators, and linked records.

Other vignettes:

Installation

Install from GitHub using pak or remotes:

# Recommended
pak::pak("noamross/airtable2")

# Alternative
remotes::install_github("noamross/airtable2")

Credentials and Authentication

Create a Personal Access Token

airtable2 uses Airtable Personal Access Tokens (PATs). OAuth is not supported; you create tokens yourself at https://airtable.com/create/tokens.

When creating a token, grant at minimum the following scopes depending on what you need:

Scope Required for
data.records:read Reading records
data.records:write Writing / upserting / deleting records
schema.bases:read Type coercion, schema inspection, air_schema()
schema.bases:write Creating bases, tables, fields
webhook:manage (Optional) Webhooks

For most workflows, enabling all scopes on the workspaces you need is simplest.

Supplying your token

There are three ways to provide your token, checked in this order:

  1. Function argument — pass .token = "patXXXXX" to any air_* function.
  2. R option — options(airtable2.token = "patXXXXX") for the session.
  3. Environment variable — AIRTABLE_API_KEY=patXXXXX in your shell or .Renviron.

The recommended approach for persistent credentials is .Renviron (or .env for multilingual projects), which keeps tokens out of your scripts:

# Add these lines to ~/.Renviron (run usethis::edit_r_environ() to open it)
AIRTABLE_API_KEY = patXXXXXXXXXXXXXX
AIRTABLE_WORKSPACE_ID = wspXXXXXXXXXXXXXX
AIRTABLE_BASE_ID = appXXXXXXXXXXXXXX # convenient if all your work is one one base

Never hardcode tokens in scripts or check them into version control.

To set validate a token interactively, use air_set_token(), which confirms the token works and stores it for the session:

air_set_token("patXXXXXXXX")
# Token set. Authenticated as "you@example.com".

Setting a default base

Most air_* functions accept base_id = NULL and will use the session default when none is supplied. Set it once per session:

air_set_base("appXXXXXXXXXXXXXX")
# Default base set to "My Base" (appXXXXXXXXXXXXXX).

The same resolution order applies to base_id:

  1. Function argument.
  2. getOption("airtable2.base_id").
  3. Sys.getenv("AIRTABLE_BASE_ID").

Checking access

Run at_sitrep() to see who you are authenticated as and which bases are accessible:

at_sitrep()
# ── Airtable situation report ───────────────────────────────────────────────
# User: you@example.com (usrXXXXXX)
# Scopes: (not exposed by this token type)
# Accessible bases: 3
#   My Base (appXXXXXX) [create]
#   Shared Base (appYYYYYY) [read]
#   ...

Finding your workspace ID

Airtable does not expose workspace IDs through its API for non-enterprise accounts. To find yours, open Airtable in a browser — the URL looks like https://airtable.com/wspXXXXXXXXXXXXXX/.... The wspXXXXXXXXXXXXXX portion is your workspace ID. You will need this for air_restore() and at_create_base().

API Limitations

Monthly call quota

Free Airtable plans cap each workspace at approximately 1,000 API calls per month. Complex operations (pagination, schema fetches for type coercion, batch writes) can each consume many calls. The package keeps a best-effort on-disk counter per workspace.

# Check how many calls you have used this month
air_api_usage()
# Airtable API usage for wspXXXXXX
# 47 API calls since start of month (2026-06-01T00:00:00Z)
# Free/team plan ceiling: ~1000 calls per workspace per month

To disable counting (e.g., in automated scripts where you track usage elsewhere):

options(airtable2.count_api = FALSE) # or AIRTABLE2_COUNT_API=false environment variable

No API deletion of bases, tables, or fields on free tier

DELETE /meta/bases/{baseId} returns 403 on free and team plans. There is no programmatic way to delete bases. Bases, tables, and most fields must be deleted via the Airtable web interface.

Checkbox fields

Airtable omits unchecked checkbox fields from API responses rather than returning FALSE. air_read() fills these in automatically using the schema, so checkbox columns always contain TRUE/FALSE with no missing values.

Computed and read-only fields

Formula, rollup, lookup, count, autoNumber, createdTime, lastModifiedTime, createdBy, and lastModifiedBy fields are read-only. air_write(), air_upsert(), and air_sync() automatically detect these from the schema and silently exclude them from write payloads, so you can safely pass a full tibble (including computed columns) and the package handles the rest.

Walkthrough

The examples below cover the full airtable2 workflow. All code uses eval = FALSE because vignettes are built without a live Airtable token. To follow along, create a demo base first (see Step 2).

Step 1 — Setup

Set your token and default base, then confirm access:

Step 2 — Create a demo base

air_demo_setup() creates a small base that exercises all of airtable2’s special types. You need your workspace ID (see the section above on finding it).

demo_base_id <- air_demo_setup(
  workspace_id = Sys.getenv("AIRTABLE_WORKSPACE_ID")
)
air_set_base(demo_base_id)

Open the demo base in a browser to see its structure and watch it update.

Step 3 — Read records

air_read() auto-paginates, fetches the table schema for type coercion, and returns a tibble. Every result includes airtable_id (Airtable’s internal record ID) and airtable_created_time.

artists <- air_read("Artists", demo_base_id)
artists
# A tibble: 15 × 10
#   airtable_id  airtable_created_time Name            Age Active Role               Disciplines  `Member Since` Email
#   <chr>        <dttm>                <chr>         <dbl> <lgl>  <chr>              <list>       <date>         <chr>
# 1 recABCDEFGH  2026-05-01 10:00:00   Zara Okonkwo     33 TRUE   Guerilla Muralist  <chr[2]>     2022-03-15     zara@bollardsforart.org
# ...

Fields with multiple values (multi-select, linked records, attachments) are returned as list-columns. See vignette("special-types", package = "airtable2") for how to work with them.

You can filter and sort at read time:

active_artists <- air_read(
  "Artists",
  demo_base_id,
  formula = "AND({Active} = TRUE, {Age} > 30)",
  sort = c(`Member Since` = "desc"),
  fields = c("Name", "Role", "Disciplines")
)

Step 4 — Write records

air_write() creates new records. Columns not present in the table generate an error by default; set add_fields = "yes" to create them automatically as singleLineText fields, or add_fields = "warn" to silently drop unknown columns.

new_artists <- data.frame(
  Name = c("Dana Reyes", "Eli Sato"),
  Age = c(28L, 34L),
  Active = c(TRUE, TRUE),
  Role = c("Guerilla Muralist", "Kinetic Sculptor")
)

ids <- air_write(new_artists, "Artists", demo_base_id)
# Created 2 records.

# Add a field that doesn't exist yet
new_artists$Notes <- "Added via API"
ids <- air_write(new_artists, "Artists", demo_base_id, add_fields = "yes")
# Creating field "Notes" in "Artists".
# Created 2 records.

Computed fields in your data frame are silently dropped before writing.

Step 5 — Upsert and sync

air_upsert() uses Airtable’s native upsert endpoint: records matching merge_on are updated; records without a match are created.

updated <- data.frame(
  Name = c("Zara Okonkwo", "Dmitri Volkov", "New Collaborator"),
  Age = c(35L, 42L, 27L)
)

result <- air_upsert(
  updated,
  "Artists",
  merge_on = "Name",
  base_id = demo_base_id
)
# Upsert complete: 1 created, 2 updated.
result$created # record IDs of new records
result$updated # record IDs of updated records

air_sync() is a more aggressive diff-based sync: it reads the current table, hashes non-key fields to detect changes, upserts only the changed rows, and optionally (but by default!) deletes rows that are absent from your local data.

desired <- data.frame(
  Name = c("Zara Okonkwo", "Dmitri Volkov"),
  Age = c(35L, 42L)
)

result <- air_sync(
  desired,
  "Artists",
  key = "Name",
  base_id = demo_base_id,
  delete_missing = TRUE
)
# Sync complete: 0 created, 2 updated, 13 deleted, 0 unchanged.

Both functions exclude computed fields automatically.

Step 6 — Join local data with an Airtable table

air_left_join() fetches a remote table and joins it with a local data frame, so you can enrich local data with Airtable fields (or vice versa) without pulling everything into memory first.

scores <- tibble::tibble(
  Name = c("Zara Okonkwo", "Dmitri Volkov", "Sun-Li Park"),
  Score = c(98, 85, 92)
)

enriched <- air_left_join(scores, "Artists", demo_base_id, by = "Name")
# Joining on "Name".

air_inner_join() and air_full_join() are also available.

Step 7 — Attachments

Download attachments to disk or as in-memory blobs:

# Download attachment metadata only (default for air_read)
df <- air_read("Projects", demo_base_id)
# df$Files is a list-column of attachment metadata

# Download files to disk
df_with_files <- air_read(
  "Projects",
  demo_base_id,
  attachments = "file",
  attachment_dir = "downloads/"
)

# Or fetch attachments for specific records
att <- air_read_attachments(
  demo_base_id,
  "Projects",
  field = "Files",
  dest = "file",
  dir = "downloads/"
)
att
# A tibble: 5 × 6
#   airtable_id filename  url         size  type            local_path
#   <chr>       <chr>     <chr>      <int>  <chr>           <chr>
# 1 recABCDEFG  report.pdf https://… 24601  application/pdf downloads/report.pdf

Upload attachments:

to_upload <- tibble::tibble(
  airtable_id = c("recABCDEFG"),
  file_path = c("new_report.pdf")
)

air_write_attachments(
  demo_base_id,
  "Projects",
  field = "Files",
  data = to_upload
)
# Uploading 1 attachment...
# Upload complete.

Attachment downloads are parallelised by default (up to 5 concurrent) using httr2::req_perform_parallel(). Disable with options(airtable2.parallel = FALSE).

Step 8 — Special types

Multi-select, linked records, collaborators, attachments, and barcodes are all returned as list-columns with custom print methods, so they display compactly in tibble output but retain their full structure.

# Read Artists - Disciplines is a list-column (air_multiselect)
artists <- air_read("Artists", demo_base_id)
artists$Disciplines
# <air_multiselect[15]>
# [1] "Mural; Street Art"    "Sculpture; Community" ...

# Flatten to a character column
air_flatten_multiselect(artists$Disciplines)
# [1] "Mural; Street Art" "Sculpture; Community" ...

# Upload with flat strings — auto-expanded (no manual air_expand_*() needed):
new_row <- data.frame(Name = "New Artist", Disciplines = "Mural; Sculpture")
air_write(new_row, "Artists", demo_base_id)

See vignette("special-types", package = "airtable2") for full coverage of all special types and their flatten/expand helpers.

Step 9 — Schema and metadata

air_schema() returns a tidy tibble of tables and fields:

schema <- air_schema(demo_base_id)
schema
# A tibble: 3 × 4
#   table_id       table_name  table_description fields
#   <chr>          <chr>       <chr>             <list>
# 1 tblXXXXXXXXXX  Artists     NA                <tibble[8 × 4]>
# 2 tblXXXXXXXXXX  Projects    NA                <tibble[7 × 4]>
# 3 tblXXXXXXXXXX  Grants      NA                <tibble[6 × 4]>

# Drill into a table's fields
schema$fields[[1]]
# A tibble: 8 × 4
#   id             name          type             description
#   <chr>          <chr>         <chr>            <chr>
# 1 fldXXXXXXXXXX  Name          singleLineText   Full name of the artist...
# ...

air_meta() returns the same information as a flat tibble (one row per field), which is easier to filter and edit:

meta <- air_meta(demo_base_id)
meta[meta$table_name == "Artists", c("field_name", "field_type")]
# A tibble: 8 × 2
#   field_name    field_type
#   <chr>         <chr>
# 1 Name          singleLineText
# 2 Age           number
# 3 Active        checkbox
# 4 Role          singleSelect
# 5 Disciplines   multipleSelects
# 6 Member Since  date
# 7 Email         email

See vignette("metadata-backup", package = "airtable2") for how to push metadata changes back to Airtable and keep bases self-documenting.

Step 10 — Backup and restore

air_dump() exports an entire base — schema plus all table data — in your choice of format:

# Full JSON backup with attachment files
air_dump(demo_base_id, dir = "backup/", attachments = "file")
# Dumping 3 tables from base "bollardsforart_demo"...
#   Reading "Artists"...
#   Reading "Projects"...
#   Reading "Grants"...
# Dump written to backup/

# Quick in-memory dump without downloading attachments
snapshot <- air_dump(demo_base_id, format = "list", attachments = "meta")
# snapshot$schema     — list of table + field definitions
# snapshot$Artists    — tibble of Artists records
# snapshot$Projects   — tibble of Projects records
# snapshot$Grants     — tibble of Grants records

air_restore() recreates a base from a dump. You must specify the workspace to restore into, or use the default one from the session.

new_base_id <- air_restore(
  "backup/",
  workspace_id = Sys.getenv("AIRTABLE_WORKSPACE_ID"),
  base_name = "Restored Base"
)
# Creating base "Restored Base"...
# Adding fields...
# Inserting records...
# Recreated 2 link fields and 0 dependent fields.
# Re-linked records in 1 table.
# Restore complete. New base ID: appYYYYYYYY.

By default (restore_linked_fields = TRUE), linked-record fields and their cell values are restored automatically. Set restore_linked_fields = FALSE to skip this step.

Note: some field types (rollup, lookup, lastModifiedTime, etc.) cannot be created via the API; air_restore() warns about them and you will need to recreate those fields manually in the web UI.

See vignette("metadata-backup", package = "airtable2") for the full backup/restore workflow.

Step 11 — Monitor API usage

Check how many calls you have used this month:

air_api_usage()
# Airtable API usage for wspXXXXXXXXXXXXXX
# 127 API calls since start of month (2026-06-01T00:00:00Z)
# Most recent call: 2026-06-04T14:22:10Z
# Free/team plan ceiling: ~1000 calls per workspace per month

The counter is per-workspace, resets at the start of each UTC month, and is stored on disk in tools::R_user_dir("airtable2", "data"). It counts best-effort (failed requests may still be counted if they reached Airtable’s servers). Opt out with options(airtable2.count_api = FALSE).

Step 12 — Teardown 

# Browse the base before cleaning up
air_browse(demo_base_id)

Bases cannot be deleted via the API on free and team plans. After a demo or experiment, delete the base manually in the Airtable web UI at https://airtable.com.

Further resources