Airtable AI Footnotes
What Airtable's AI features do with your data, as Airtable publishes it. Checked October 31, 2026. Airtable in the Directory.
Agent access
Airtable's MCP server at https://mcp.airtable.com/mcp is available to all users on all plans, at no extra charge. Enterprise Scale admins can block MCP access, block third-party API integrations, or allowlist specific MCP clients by OAuth client ID. source, source 2, source 3, source 4, source 5 · checked October 31, 2026
Omni (AI assistant)
Airtable AI on all plans
| On by default? | Airtable AI is on by default for all plans, including legacy plans, except for EU and GDPR customers, who turn it on by choice. The admin panel's Omni setting is on by default and off by default in EUDR countries. source, source 2, source 3 · checked October 31, 2026 |
|---|---|
| How an admin turns it off | Business and Enterprise Scale admins: admin panel > Settings > AI settings, where the Airtable AI switch turns off AI in all organization-owned workspaces and the Enabling Omni setting turns off Omni. Workspace owners: account page > Workspace settings > Workspace AI permissions. Base Owners and Creators can turn Omni off per base, which collaborators can reverse. source, source 2, source 3, source 4 · checked October 31, 2026 |
| Trains on customer data? | Airtable does not use, or let its AI providers use, Input, Output or other Customer Data to train generative AI models. source · checked October 31, 2026 |
| Retention | Enterprise Scale and Business: AI providers do not log Input or Output for human review or retain it. Free, Team, Pro and Plus: AI providers may keep Input and Output for up to 30 days for safety and compliance moderation, and longer where moderation confirms prohibited content. On all plans, providers may keep usage metadata for billing, safety and compliance. source, source 2, source 3, source 4 · checked October 31, 2026 |
| Where data is processed | The sub-processor list gives Global as the processing location for AWS, Google and OpenAI, and USA for Anthropic and IBM. Airtable's data residency page (US, Europe, Australia on Enterprise Scale) covers data at rest and does not mention AI processing. source, source 2 · checked October 31, 2026 |
| Model providers | Omni needs OpenAI and/or Anthropic (Amazon Bedrock) models enabled. The AI sub-processors are AWS, Anthropic, Google, IBM and OpenAI as model providers, with Castle Global (moderation, Enterprise and Business plans), Parallel Web Systems (web search), Reducto (data extraction) and Vercel (code execution). Business and Enterprise Scale admins can choose which providers are enabled. source, source 2, source 3 · checked October 31, 2026 |
| Price | Included. Building apps and agents with Omni uses no AI credits; asking questions about base data costs about 10 credits. Monthly credits are included per plan: 500 per editor on Free, 15,000 per paid user on Team, 20,000 on Business and 25,000 on Enterprise Scale at list price. source, source 2, source 3 · checked October 31, 2026 |
Field agents (AI in fields)
Airtable AI on all paid plans
| On by default? | Not published as of October 31, 2026 |
|---|---|
| How an admin turns it off | Controlled by the same Airtable AI switches (admin panel > Settings > AI settings, or Workspace AI permissions); AI credits and workspace-level AI are required. Admins can turn off internet access for AI and web prototype generation by field agents. source, source 2, source 3 · checked October 31, 2026 |
| Trains on customer data? | Airtable does not use, or let its AI providers use, Customer Data to train generative AI models. source · checked October 31, 2026 |
| Retention | Enterprise Scale and Business: AI providers do not retain Input or Output. Free, Team, Pro and Plus: up to 30 days for moderation. source, source 2 · checked October 31, 2026 |
| Where data is processed | Not published as of October 31, 2026 |
| Model providers | Users pick a model from those their plan and admin settings allow. The AI sub-processor list (updated October 30, 2026) names AWS, Anthropic, Google, IBM and OpenAI as AI model providers. The Third-Party AI Providers page lists Meta Llama and Mistral terms for AWS, and Meta Llama terms for IBM watsonx. source, source 2, source 3, source 4, source 5 · checked October 31, 2026 |
| Price | Consumption in AI credits, varying with input, model and output: about 1 credit to categorize feedback, 10 for a web search and 200 to analyze a 10-page contract. Extra credits cost 20 USD a month per 10,000 (200 USD a year). source, source 2, source 3 · checked October 31, 2026 |
Canvas
Airtable AI (open beta) for organization-owned bases
| On by default? | An admin turns on Enable Canvas in admin panel > Settings > AI settings, after Airtable AI is on and Anthropic or Amazon Bedrock is allowed. source · checked October 31, 2026 |
|---|---|
| How an admin turns it off | Admin panel > Settings > AI settings > Enable Canvas; on Enterprise Hub each org unit has its own switch. Workspace owners who turn off AI for their workspace also turn off Canvas there. source · checked October 31, 2026 |
| Trains on customer data? | Not published as of October 31, 2026 |
| Retention | Canvas stores data to run projects, and Airtable's normal data retention rules do not cover it. source · checked October 31, 2026 |
| Where data is processed | Not available to organizations with regional data residency, HIPAA compliance or Enterprise Key Management. source · checked October 31, 2026 |
| Model providers | Anthropic or Amazon Bedrock must be allowed as a model provider. source · checked October 31, 2026 |
| Price | Not published as of October 31, 2026 |
Compliance
- HIPAA: Enterprise Scale only, with a signed Health Information Exhibit (which includes the BAA). Airtable AI is available to these customers once they accept the AI Terms, including the Supplemental AI Terms for Health Information. Several AI sub-processors (Castle Global, Parallel Web Systems, Reducto, Vercel) do not apply to HIPAA-enabled accounts, and Canvas is not available to them. source, source 2, source 3 · checked October 31, 2026
Not published by Airtable as of October 31, 2026
- Region where LLM inference runs for Airtable AI, beyond the sub-processor list's Global and USA entries
- Whether data residency in Europe or Australia keeps AI processing in that region
- A release stage for the Airtable MCP server
- Whether Omni connects to outside MCP servers
- EU AI Act statement (Airtable is not on the European Commission GPAI Code of Practice or AI Pact signatory lists on the check date)
- FedRAMP authorization
Changes
No changes recorded since the first check on October 31, 2026.
How this page is made, and how to send a correction: method.