An Airtable MCP that treats your bases like the databases they are
Connect one Personal Access Token and Airtable bases become workflow steps: list, create, update, upsert, and delete records in batches, read table schemas, manage fields, and keep comments and webhooks under automation control. The one API operation that silently wipes fields is deliberately not shipped.
What the Airtable MCP connector does
The Airtable connector for Mattering Automations is 22 nodes across the Web API: account metadata (who am I, list bases, get base schema), records (list with formula filters, views, and sorts; get; batch create, update, upsert, and delete of up to 10 records per call), schema writes (create and update tables and fields), record comments, and webhook subscription management (list, create, refresh, delete, and fetch payloads).
Every node speaks MCP, the open protocol for connecting AI assistants to tools, so Claude, ChatGPT, and any MCP-capable assistant can use them, alongside Matter, the AI assistant built into the platform. An assistant can answer which records match a formula, read a base's schema to understand what lives where, and file new records, without anyone opening Airtable.
Because Mattering is a work platform rather than middleware, Airtable data lands somewhere useful by default: a scheduled records sweep can open tasks in Mattering Productivity for rows that need human attention, and finished work can flow back as record updates, all on one canvas next to the rest of the work.
Safety is modeled honestly. Record updates ship as PATCH, which only touches the fields you send; Airtable's PUT-style replace, which silently clears every field you omit, is deliberately not a node. Deletes are flagged destructive. And because the API offers no way to delete a table or a field, the nodes that create them warn that they cannot clean up after themselves.
One connection, three surfaces
- Visual automations. Drag Airtable nodes onto the canvas and wire them to triggers, conditions, and other apps.
- AI agents, yours or ours. These nodes speak MCP, the open protocol, so Claude, ChatGPT, and any MCP-capable assistant can use them, and Matter, the agent built into the platform, uses them natively.
- The whole platform. Combine with native tasks, projects, and other connectors, no middleware in between.
Every Airtable node, ready to drop on the canvas
22 nodes, generated from the connector itself so this list always matches what ships. Expand a category to browse.
Read and query nodes 8
Airtable: Get Base Schema
Return the full schema of a base: every table with its id, name, primary field, fields (id, name, type, options) and views. This is where table ids (tblXXX) and field ids (fldXXX) come from. Requires the schema.bases:read scope.
Airtable: Get Record
Fetch one record by id, returning its fields and created time. Empty cell values (empty string, false, empty array) are omitted from the response. Attachment cells return URL objects whose URLs expire 2 hours after the response, so download promptly rather than storing the URL. Requires the data.records:read scope.
Airtable: List Bases
List every base the token can access, with the permission level on each. This is where the base ids (appXXX) used by every other Airtable node come from. Returns up to 1,000 bases per page with an offset cursor. Requires the schema.bases:read scope; a base missing from the token's resource list will not appear.
Airtable: List Comments
List the comments on a record, newest first, with author, text, reactions and thread parent ids. Requires the data.recordComments:read scope.
Airtable: List Records
List records from a table, with filtering (filterByFormula), a view, sorting, field selection and offset pagination (up to 100 records per page; data.offset carries the cursor for the next page). A read despite the POST: this is Airtable's documented POST /listRecords variant of the list endpoint, used so long formulas and sort arrays travel in the JSON body instead of the query string. Requires the data.records:read scope. Airtable allows 5 requests per second per base; a 429 means wait 30 seconds.
Airtable: List Webhook Payloads
Fetch the change payloads a webhook has accumulated, in transaction order, with a cursor for the next call and a mightHaveMore flag. Side effect worth knowing: calling this on an active webhook resets its 7-day expiration clock. Requires the webhook:manage scope.
Airtable: List Webhooks
List the webhooks registered on a base with their specification, notification URL, enabled state, expiration time and payload cursor. Requires the webhook:manage scope.
Airtable: Who Am I
Return the user id behind the token (plus email when the token has the user.email:read scope). Requires NO scopes, so it is the cleanest credential health check: 200 means the token is valid, 401 means it is not. Note a 200 here does not prove data access; scopes and the token's base list gate every other node.
Action nodes 14
Airtable: Create Comment
Post a comment on a record, optionally as a threaded reply. Mention a collaborator inline with @[usrXXX] and Airtable notifies them. Requires the data.recordComments:write scope and base commenter access.
Airtable: Create Field
Add a new field (column) to a table. Additive, but note the Web API has NO delete-field endpoint, so this cannot be undone from an automation; removal is manual in the Airtable UI. Requires the schema.bases:write scope and base creator access.
Airtable: Create Records
Create up to 10 records in one call. Each record is a {"fields": {...}} object keyed by field name or id. To attach a file, set an attachment field to [{"url": "https://example.com/file.pdf"}] and Airtable fetches it server-side. Requires the data.records:write scope and base editor access. Airtable allows 5 requests per second per base; a 429 means wait 30 seconds.
Airtable: Create Table
Create a new table in a base, defining its fields up front (the first field becomes the primary field). Additive, but note the Web API has NO delete-table endpoint, so this cannot be undone from an automation; removal is manual in the Airtable UI. Requires the schema.bases:write scope and base creator access.
Airtable: Create Webhook
Register a webhook on a base. Management only: this does NOT create an inbound trigger on this platform; notificationUrl must be an endpoint you operate (or omit it and poll List Webhook Payloads). PAT-created webhooks expire after 7 days unless refreshed. The response's macSecretBase64 (for verifying pings) is shown ONLY here and cannot be retrieved again, so persist it immediately. Limit 10 webhooks per base. Requires the webhook:manage scope and base creator access.
Airtable: Delete Comment
Permanently delete a comment from a record. DESTRUCTIVE: there is no undo. Non-admin tokens can only delete comments their own user created. Requires the data.recordComments:write scope and base commenter access.
Airtable: Delete Records
Permanently delete up to 10 records by id in one call. DESTRUCTIVE: deleted records are only recoverable through Airtable's own trash/snapshot features, not through the API. A single id in the list deletes one record. Requires the data.records:write scope and base editor access.
Airtable: Delete Webhook
Permanently delete a webhook from a base. DESTRUCTIVE: the webhook, its undelivered payloads and its MAC secret are gone; a replacement gets a new id and secret. Requires the webhook:manage scope and base creator access.
Airtable: Refresh Webhook
Extend an active webhook's life by 7 days from now and return the new expiration time. Schedule this on a cadence shorter than 7 days to keep a webhook alive indefinitely. Requires the webhook:manage scope and base creator access.
Airtable: Update Comment
Edit a comment's text. The API only lets a token edit comments its own user created. Requires the data.recordComments:write scope and base commenter access.
Airtable: Update Field
Update a field's name, description, or type-specific options (send at least one). The field's TYPE cannot be changed through the API. Requires the schema.bases:write scope and base creator access.
Airtable: Update Records
Partially update up to 10 records in one call: only the fields you send change, everything else is left alone (PATCH). Clear a single field by sending null for it. Airtable's PUT variant, which wipes every field you do NOT send, is deliberately not shipped. Requires the data.records:write scope and base editor access. Airtable allows 5 requests per second per base; a 429 means wait 30 seconds.
Airtable: Update Table
Rename a table or change its description (send at least one of the two). Returns the full updated table schema. Requires the schema.bases:write scope and base creator access.
Airtable: Upsert Records
Update-or-create up to 10 records keyed on your own external id fields (Airtable's documented performUpsert mode of the PATCH update endpoint). Records without an id are matched on fieldsToMergeOn: zero matches creates, one match updates, several matches fails the request. The response lists createdRecords and updatedRecords ids separately. Requires the data.records:write scope and base editor access. Airtable allows 5 requests per second per base; a 429 means wait 30 seconds.
Connecting Airtable takes one credential
Airtable issues Personal Access Tokens from the Builder Hub (airtable.com/create/tokens). A token carries two things: scopes (like data.records:read or schema.bases:write) and a list of bases or workspaces it can touch. Both are chosen at creation, and both matter: a valid token that lacks a scope or was never granted a base returns 403s that look like permission bugs but are token configuration.
In Airtable's Builder Hub, create a Personal Access Token with the scopes your automations need, and grant it the specific bases or workspaces they will touch.
In Mattering Automations, open Connectors, click Add Connector, choose Airtable, and create a credential with that token.
Click Test Connection. Mattering calls Airtable's who-am-i endpoint, which needs no scopes at all, so the test cleanly answers whether the token itself is valid.
All 22 Airtable nodes appear in the canvas palette. Run Airtable: List Bases next: it confirms the token can actually see data, and returns the base ids every record node asks for.
Common uses for the Airtable MCP
Records in and out, ten at a time
Airtable: List Records supports filterByFormula, views, sorts, and pagination, and posts through Airtable's documented body variant so complex formulas travel safely. Airtable: Create Records, Airtable: Update Records, and Airtable: Upsert Records write up to 10 records per call, and upsert matches on the fields you choose, which makes syncs idempotent.
A base an agent can understand
Airtable: Get Base Schema returns every table, field, and field type in a base. An assistant reads it once and then knows what the base contains, which fields are computed, and what a valid record looks like, before it writes anything.
Two-way flow with Mattering work
A scheduled Airtable: List Records sweep with a formula like NOT({Status} = 'Done') can open tasks in Mattering Productivity for rows needing attention, and closing the task can drive Airtable: Update Records to mark the row. Neither side needs to know the other exists; the canvas is the glue.
Schema managed from the canvas
Airtable: Create Table, Airtable: Create Field, and their update twins let provisioning workflows build out a base's structure. Both create nodes warn honestly that the API has no delete for tables or fields, so what a workflow creates, only a person in the Airtable UI can remove.
Comments where the discussion is
Airtable: List Comments, Airtable: Create Comment, and the update and delete nodes keep record-level discussion automatable: a workflow that flags a record can say why, in the place a person will look.
Webhooks managed, honestly labeled
Airtable: Create Webhook, Airtable: List Webhooks, Airtable: Refresh Webhook, Airtable: Delete Webhook, and Airtable: List Webhook Payloads manage Airtable's webhook subscriptions and read their change payloads. There is no inbound Airtable trigger in this connector; the subscriptions can point at any URL you operate, and a scheduled payload fetch covers polling.
Example: A leads base that files its own follow-ups
Built on the Automations canvas in a few minutes, or described in plain language to Matter and built for you.
Trigger: a schedule runs each morning.
Airtable: List Records pulls rows from the leads base with filterByFormula matching leads marked for follow-up.
A task is created in Mattering Productivity for each lead, carrying the record id and the fields a person needs.
When the task completes, Airtable: Update Records patches just the status field on the matching record.
Airtable: Create Comment writes what happened onto the record, so the trail lives where the data lives.
Airtable MCP questions, answered
What Airtable operations are supported?
22 nodes: account metadata (who am I, list bases, get base schema), records (list with formulas, views, and sorts; get; batch create, update, upsert, delete), table and field creation and updates, record comments, and webhook management including payload fetching. Deliberately not shipped: Airtable's PUT-style record replace (it silently clears omitted fields; PATCH plus explicit nulls covers the same ground safely), single-record update and delete (strict subsets of the batch nodes), base creation (needs a workspace id no public endpoint can list, and cannot be undone), and the base64 attachment upload endpoint.
Does this work with Claude, ChatGPT, or other AI assistants?
Yes. Mattering exposes these nodes over MCP, an open protocol for connecting AI assistants to tools, so any MCP-capable assistant, including Claude and ChatGPT, can connect to your workspace and use them. Matter, the assistant built into the platform, uses the same nodes natively.
My token tests fine but nodes return 403. Why?
That is the signature Airtable quirk. A Personal Access Token carries scopes AND a per-base access list, both fixed at creation. The connection test uses the who-am-i endpoint, which needs no scopes, so it proves the token is valid without proving it can reach your data. A 403 on a real node almost always means the token lacks a scope (like data.records:write) or was never granted that base. Edit the token's configuration in Airtable's Builder Hub, or mint a new one.
How many records can one call touch?
Ten per create, update, upsert, or delete call, per Airtable's API. For larger volumes, loop the batch nodes; Airtable's rate limit is 5 requests per second per base, and after a 429 the API expects a 30-second pause, which the node surfaces so a workflow can wait and retry. A sweep of a few hundred records comfortably fits inside these limits.
What makes upsert different from create or update?
Airtable: Upsert Records matches incoming records against existing ones using the merge fields you name: matches get updated, misses get created, and the response reports which ids were created versus updated. That makes recurring syncs idempotent, so running the same sync twice does not duplicate rows.
Can an Airtable change trigger a workflow?
Not as a trigger node in this connector. The webhook nodes manage subscriptions and fetch their change payloads rather than receiving events, so there is no inbound Airtable trigger. The honest patterns today: point a subscription's notifications at a URL you operate, or run Airtable: List Webhook Payloads on a schedule and act on the changes it returns. Note that Airtable webhooks expire and need Airtable: Refresh Webhook to stay alive.
How do attachments work?
As URLs in both directions. Records with attachment fields return attachment objects whose download URLs are valid for about two hours, so downstream steps should use them promptly or re-fetch the record. To add an attachment, include a publicly reachable url in the field value on create or update and Airtable ingests it server-side. The base64 upload endpoint is deliberately not shipped, so raw file bytes never pass through workflow payloads.
Why do the table and field creation nodes carry a warning?
Because the Web API has no delete for tables or fields. A workflow can create both, but only a person in the Airtable UI can remove them, so a misconfigured provisioning loop leaves structure behind that automation cannot clean up. The nodes say this in their descriptions; treat schema writes as one-way and test them against a scratch base first.
Put Airtable to work inside your workspace
Connect once, then use every node in visual automations and AI agent workflows across the Mattering platform.
All product names, logos, and brands are property of their respective owners and are used for identification purposes only. Use of these names, logos, and brands does not imply endorsement or affiliation. Airtable is a product of its respective owner; Mattering is an independent platform.