🍂 Fall Release: Omni 2.0, Airtable Canvas, a bigger MCP, agent audit logs, and upcoming agent permissions

Hi everyone, Isabella here from the Airtable team 👋

We just wrapped the keynote at Airspace London, where we unveiled our Fall Release, and I wanted to bring the highlights straight here. A lot of what shipped today was shaped by conversations in this community, so I’m really excited to share with you the new products and features that AIrtable has been working on this year!

The big idea

We’re all too familiar with one particular agentic refrain “You’re right, that was a mistake.” But it isn’t because the agent isn’t smart enough. It’s almost always because it was working from data that was incomplete, out of date, or scattered across five tools. It lacked context, essentially.

You know your team’s most complete, structured, live data lives in your bases. This release is all about letting agents work from that same source of truth, with you deciding what they can touch and seeing everything they did.

Here’s what’s new 👇

🤖 Omni 2.0 (available now on all plans)

Some of you have already been taking the Omni upgrade for a spin (thank you for all the feedback in this thread!). Omni 2.0 is faster, smarter, and can take on much bigger jobs. Hand it your spreadsheets, PDFs, and images, however messy, and it’ll build the data model, import your data, and set up automations and AI fields in one pass while you watch. It can now:

• Set up Field Agents that summarize, classify, or enrich records
• Configure complex automations from plain language
• Write scripts and drop them into automations where no built-in action exists
• Model your business with tables, fields, and linked records

Open it from any base to try it if you haven't already.

🎨 Airtable Canvas (open beta coming this week)

This is the one we can’t wait to see you get your hands on. Simply describe what you want in plain language and Canvas will build you a custom app, dashboard, or presentation on top of your live data, all with your own branding. No dragging and dropping required.

Because Airtable Canvas runs on your base, it already respects all of the predetermined permissions and sharing settings, and it’ll propose a plan and ask for your sign-off before adding back any tables or fields. One base can power an app for the whole team, a dashboard for your CMO, a Monday-morning one-pager, and a weekly deck that updates itself.

Join the waitlist here.

Heads up: Airtable Canvas usage requires that AI, AI Labs ands Anthropic as an AI provider is enabled org-wide and isn’t yet available for HIPAA-compliant or non-US Data organizations.

🔌 Airtable MCP: more power, more connections

We launched Airtable MCP earlier this year, and it now connects with even more of your favorite tools like Claude, ChatGPT, Gemini, and Cursor (and soon Copilot and Muse), so you can read and write records, submit forms, and build automations and interfaces without leaving your AI tool of choice. Admins can turn MCP on or off org-wide, approve which tools people can use, and limit each tool to the bases it needs. If you missed it, here’s our earlier post on building automations through MCP.

🔍 Agent audit logs (available now for account admins)

More agents in your bases means you need to see exactly who did what. The audit log now:

• Tells you whether a person or an agent took an action, and in which base. Omni’s actions are attributed directly, and third-party AI tools trace back to the integration and the person who connected it.
• Records in-base events like creating a table, updating an interface page, or deleting a comment (rolling out now on Enterprise plans)
• Can be filtered by actor, base, event type, org unit, and date, then exported to CSV or JSON or streamed through the audit API into your security tools

🔜 Coming soon

• Agent permissions: choose exactly what an agent can read or write, down to the field or record. Let an agent work through customer feedback without ever seeing contact info. And agents never get more access than the person who connected them.
• Bring your own agent (and keys): run your organization’s own autonomous agents on your bases, with the same permission controls and visibility.

Over to you 💬

We’d genuinely love to hear from you in the replies. Which of these are you trying first? What are you itching to build with Canvas? And for the admins here: what would you need from agent permissions before you’d let an agent loose on a production base? The product team will be reading. And if you build something cool, show it off in Show and Tell!

For the full details, check out the Fall Release blog post here. And watch this space for our full recap of Airspace London!

Isabella