Databases roundtable recap (Nov 18)

Thanks to everyone who joined yesterday’s roundtable, and to everyone who dropped questions into Dory. @Nathan_Penner (PM, Superhuman Docs) walked through where Databases stands and what’s coming. Hayden, who recently joined the team as a PM, also jumped in.

Here’s the recap for anyone who missed it.

The short version

  • The private beta is expanding. It’s open to Team, Business, and Enterprise workspaces. Enroll here. General beta access opens in December.
  • It’s built for scale. Data loads on demand, with up to 1M rows per table (2M per database) on Enterprise. Connected docs update in real time, with no sync buttons and no waiting.
  • AI and MCP support are coming in the next couple of weeks. That includes native AI in databases and MCP connections for tools like Claude, ChatGPT, and Cursor.
  • Migration is manual for now. “Copy to database” makes a copy of your table, and then you rebuild your doc around it.

What you can do today

  • Create a database from scratch, import a CSV, or copy an existing table from a doc
  • Work with multiple connected tables, relations, roll-ups, and formulas
  • Use granular permissions, so people can edit rows without changing columns
  • Restore deleted rows and columns from the trash
  • See which docs use a database and remove their access
  • Show live, full-fidelity views in docs, and reference database data in Canvas formulas
  • Run a pre-migration checker that flags unsupported columns before you copy data over

Roadmap

Phase Timing What’s included
1. Working with data directly Now Central, reusable datasets and simple-to-moderate trackers, plus AI and MCP in the coming weeks
2. Richer processes Early 2027, shipping incrementally Automations, buttons, conditional formatting, custom layouts, subtables, AI Views integration, custom agents, doc-level annotation columns, and in-place CrossDoc migration
3. Advanced needs Later External data via Packs, rich collaborative Canvas columns, sharing partial subsets of data, and full CrossDoc replacement

Your questions, answered

@Tracy_Horner: How do I update a database from my existing docs, like HR forms that use buttons to write to a master table?
Databases don’t support buttons and automations yet. They’re a focus for early 2027, rolling out as a series of launches. If your workflow depends on them, check the roadmap against your needs. “Copy to database” runs a checker that flags unsupported columns. Keep in mind it makes a copy, so you’d have two sources of truth until you rebuild the doc.

@Thomas_Oldroyd: How can complex CrossDocs be replaced by databases?
It’s a two-step process. First, move your source data into a database. The “CrossDoc Bridge” lets you re-parent your child docs so they point to the new source of truth, still using CrossDoc’s slower syncs. Second, replace those CrossDoc views with native database views, which are real-time and faster. You can do that manually now. An automatic conversion tool is planned for a later milestone.

@TJ_Beatrice_SCPB : Beyond performance, how are databases different from the new table pages?
The UI is intentionally similar. The big differences are:

  • Efficient, real-time reuse of one dataset across many docs
  • Better scale and performance (the rule of thumb is 1,000+ rows)
  • Independent permissions, so people can edit rows but not delete columns
  • Trash and restore, plus visibility into where data is used

@Meg_Tazelaar: What’s the long-term plan if we outgrow 1–2M rows? Should we pipe data elsewhere, like Databricks?
There’s no specific recommendation yet. Nathan said bringing in external data is planned further out, but it’s too early to say whether it would carry the same row limits. Searching or referencing external data is a different shape of problem than syncing it all, and it may be more feasible. This one is only partly answered, so I’d love to hear how others are thinking about it.

@Bill_French : Will AI Views work with large datasets?
Yes, that’s the goal. Like native building blocks in docs, a view would show a subset of rows while aggregate calculations and dashboards run across the full dataset. The team is still planning where the boundaries are.

@William : What are the limits on lower plans?

  • Enterprise: 1M rows per table, 2M per database
  • Team/Business: 200k rows per table, 500k per database

@K.T_Wiegman : Is Team the same as Pro?
No. Team is the plan that will be supported. Pro is a level below it and initially won’t have databases.

@Tim_Richardson1: Will Docs and Databases eventually converge?**
The goal is to converge as much of the feature set as possible, especially the UI and logic layer (views, column types, automations, custom layouts). The core difference is that a database is an independent data source with better scale, performance, and reuse.

@Hugo_Assuncao1 : Will native doc tables get a hard row limit?
Eventually, but not soon. Once there’s full feature parity and a seamless, automatic way to convert tables, the plan is to limit native tables to around 1,000 rows. Nathan also said CrossDoc will eventually be phased out. There are no plans to force changes to existing docs, CrossDoc syncs, or tables over 1,000 rows. The change would only apply to new content.

@Bruce_Lieberthal : Can I edit a database schema from the docs it appears in?
Not right now. You make schema changes in the database itself. The team wants feedback on this, since there’s a balance between protecting shared data and making things efficient for data owners.

Answered in Dory (during or after the call)

We didn’t get to every question live, but Hayden and Nathan answered many of them in Dory. Here’s what we have.

Packs and external data

@Thomas_Schulz : Are Packs available directly in databases? We want to use one to propagate/update tables.
Hayden: Packs for importing and syncing data into a database are coming next year, but won’t be in the initial launch.

@Karina_Mikhli: When will Pack tables be available in databases?
Nathan: They’re on the roadmap but not supported yet.

Limits and performance

@William: Is 1 million rows per database, or per table?
On Enterprise it’s 1M rows per table and 2M total across all tables in a database (Nathan). Team and Business are 200k per table and 500k per database, as covered on the call.

@Thomas_Oldroyd: Will the doc size limits (125 MB and 325 MB) be upgraded?**
Nathan: Views of databases in docs don’t count toward those limits. If you use databases as the source for large datasets, hitting them should be rare.

@Chris_Strom : How stable is “Copy to Database”? It’s always failed on our larger docs.
Nathan: The team has made several updates, including a key fix in the past week. Try it again, and let them know if anything fails so they can investigate.

References and relations

@Tim_Richardson1 : Can databases reference other databases, or only tables within the same database? What about two databases synced into the same doc?
Hayden: References to tables in other databases will come later. Initially, references are to tables within the same database. But if two databases are synced into the same doc, you can create reference columns in a doc table pointing to tables from different databases.

Anonymous: Can you reference databases in tables in docs?
Nathan: Yes. The follow-up (how to access the actual values from the referenced row, not just the relation column) is still open. Nathan asked them to share a screenshot in the beta Slack channel or through the feedback form.

Removing connections

@Hugo_Assuncao1 : Does removing a connection delete the database view in the doc? Is managing that a Doc Maker-only action?
Nathan: Removing a connection breaks existing views, so they show an error. If you re-establish the connection, they show data again. (The Doc Maker part wasn’t addressed.)

Permissions and trash

@William : Can you restrict access to one specific table so users see some tables but not others?
Hayden: Not yet. Finer-grained permissions within a database are coming. Hayden asked for a list of what you’d like beyond table-specific view permissions, so reply in the thread if yo@Meg have one.

@Meg_Tazelaar: Would “edit rows but not configure columns” permissions be respected by MCP? Can we set a default permission for new users?
Hayden: MCP takes on the permissions of the credentials used to connect to the database. Defaults will work much like Docs today, where you set default share permissions in the share modal.

@Thomas_Oldroyd: Can individual users manage Trash, or only admins?
Hayden: Users with Full Access can manage all trash items. Other users can manage their own items, for example, to undo something they did or fix an issue from a connected MCP client.

API, MCP, and automation

@TJ_Beatrice_SCPB: How will API access work? Will it use the standard setup we’ll build, like tables do today?
Hayden: We’ll build out the main functionality after the open beta. MCP is launching sooner, so an AI can do these things for you. Hayden added that the goal is for databases to be the backend for any app you build, whether in Docs or an AI platform.
@Bill_French asked whether MCP will be in the beta, and the answer is yes, it’s coming soon.

@Thomas_Oldroyd: What triggers are supported? Can they be scheduled, or can external agents trigger them?
Nathan: Agents will support scheduling by time, row modified, or events from other apps. Automations will support time-based, row-based, and webhook triggers.

CrossDoc

@K.T_Wiegman: Is the goal to retire CrossDoc entirely, or will it keep working if we don’t upgrade a doc?
Nathan: CrossDoc will keep working in existing docs. Eventually, we’ll disable creating new CrossDoc connections.

Sharing your experience

@Bill_French : I’m exploring databases as a foundation for systems of record. Guidance on what we can share openly?
Nathan: Exciting, and the team would love to see what you explore. Private beta usage is under NDA, but they’d love for you to share your experience once the open beta begins.

Keep the conversation going!

Add anything else on your mind below, and tag me with anything you’d like us to pass along to the product team.

As always, thank you for being so present and for asking such great questions during the roundtable. Can’t wait for the next one!

2 Likes