Connect Databricks to Superhuman Docs and Superhuman Go

Superhuman Connectors let you bring the tools your organization relies on directly into Superhuman Docs and Superhuman Go. With the Databricks connector, you can pull data from your Databricks workspace directly into Superhuman Docs and ask questions about it in plain language through Superhuman Go, our AI chat agent — including natural-language Q&A against your Genie Spaces and Unity Catalog data. This ensures that reporting and decision-making are consistently based on your real-time data, allowing your team to derive insights without needing to export spreadsheets or switch tools.

Note: The Databricks connector is only available to customers on our Enterprise tier. You must be an Enterprise admin to complete the setup and configuration steps outlined in this article.

Configure the Databricks connector for your organization

Before the Databricks connector can be used within your organization, an admin must complete a few initial configuration steps. This process has two main parts: creating a service principal in Databricks and configuring the connector in Superhuman.

Part 1: Create an OAuth application in Databricks

For this part, you’ll jump over to Databricks. Note that you will need to be a Databricks account admin to follow these steps:

1. Log in to the Databricks account console (for example, accounts.cloud.databricks.com) as an account admin.
2. Click the Settings icon in the sidebar, open the App connections tab, and click Add connection.
3. Give the application a name — for instance, “Superhuman Databricks connector”.
4. Under Redirect URLs, add these:
   • https://docs.superhuman.com/packsAuth/oauth2/26326
   • https://coda.io/packsAuth/oauth2/26326
5. Under Access scopes, grant the scopes Superhuman requests:
   • genie — natural-language Q&A against your Genie Spaces.
   • unity-catalog — read access to Unity Catalog data.
   • offline_access — lets Superhuman refresh access without prompting each person to sign in again.
6. Leave the access token TTL and refresh token TTL at their defaults (60 minutes and 10080 minutes) unless your organization requires different values.
7. Select Generate a client secret, then save the following pieces of info:
   • Client ID — generated for the application.
   • Client Secret — shown once when you generate it, so save it securely before closing the dialog.

Unlike a service-principal setup, you do not grant data permissions to this application. The application only establishes trust between Databricks and Superhuman; each person's access is governed by their Databricks permissions. Make sure the people who will use the connector have:

• USE CATALOG, USE SCHEMA, and SELECT on the catalogs and schemas that members should be able to read from in Unity Catalog.
• Access to any Genie Spaces they should be able to ask questions about.

Keep the Client ID and Client Secret handy as you go to your Superhuman account to complete the configuration.

Part 2: Configure the connector in Superhuman

With the necessary info from Databricks, you can now complete the configuration in Superhuman:

1. Go to the Connectors page.
2. In the search field in the upper-right corner, enter Databricks, then click on the Databricks connector to start configuring it.

3. On the page that opens, click on the menu next to Connector access and select Custom.

   

4. In the window that appears, click Confirm.

5. In the Configurations section, click Edit to change the existing configuration, or click + Add Configuration below to create a new one.

   

6. Enter or edit the configuration name, then under Groups with access, choose whether the configuration applies to everyone or to selected groups.
7. In the Policy section, for the Connection endpoint URL, enter your Databricks workspace URL — the URL you use to access Databricks in the browser (for example, https://your-workspace.cloud.databricks.com).
8. Under Read permissions, select Allow, Always ask, or Deny.
9. Under Write permissions, also select Allow, Always ask, or Deny.
10. Optionally, check the box next to Allow only private connections to write to Databricks.
11. Optionally, check the box next to Doc sharing to enforce configuration permissions.
12. Next, in the OAuth app credentials section, enter your Client ID and Client secret copied from your Databricks workspace.
13. For the Authorization URL, enter your URL in this format: https://{your-workspace-url}/oidc/v1/authorize
14. For the Token URL, enter your URL in this format: https://{your-workspace-url}/oidc/v1/token
15. Click Save configuration.

Note: Superhuman requests the genie, unity-catalog, and offline_access scopes from Databricks and uses PKCE to secure the exchange — there’s no separate scope selection step in Superhuman. These scopes must be granted to the application in the Databricks account console as described earlier, or sign-in will be rejected.

Try out Databricks in Superhuman Docs

Unlike connectors backed by a single shared credential, Databricks asks each person to sign in with their own account. After an admin completes the configuration above, every member who wants to use the connector does this once:

To do that, follow these steps:

1. Type /Databricks in your doc, and select the Databricks connector from the options.
2. In the window that appears, click Select configuration, choose a configuration from the dropdown, and click Select.
3. Click Connect account.
4. Sign in to Databricks and approve the requested access.

That’s it — the connection belongs to that person from now on.

What this means for your organization:

• Access follows the person. Each member can query only the catalogs, schemas, and Genie Spaces their Databricks permissions allow. Two people asking the same question may get different results, which is expected.
• Activity is attributed to the individual. Queries and Genie conversations appear under that person’s identity in Databricks query history and audit logs, rather than under a shared identity.
• Databricks access is a prerequisite. Someone without a Databricks account, or without permissions to the relevant data, can’t use the connector until that access is granted in Databricks.
• Revoking access works as you’d expect. When someone’s Databricks access is removed, their Superhuman connection stops working too — no separate cleanup needed in Superhuman.

Tap into Databricks data in Superhuman Go

Superhuman Go is our AI chat agent, available in the Superhuman browser extension for Chrome and Edge, and in the AI editor at app.grammarly.com. Once the Databricks connector is configured, Go can use Databricks’ own Genie — natural-language Q&A over your Genie Spaces and Unity Catalog data — to help you work with your data in plain language, without writing SQL or leaving the page you’re on.

To start Go in the browser, hover over the right edge of your browser window to reveal the Superhuman tab, then click it to open the chat. You can then ask Go to work with your Databricks data. For example, you can:

• Ask questions in plain language: Query your Genie Spaces without writing SQL — for example, “What was total revenue by region last quarter?” — and get an answer grounded in your actual Unity Catalog data.
• Keep a conversation going: Ask follow-up questions, and Go will retain context from earlier turns in the same conversation, just like a normal back-and-forth.
• Pull numbers into your work: Drop specific figures or answers straight into the doc, email, or message you’re drafting.
• Generate content grounded in your data: Turn a Genie answer into a summary, status update, or report without exporting anything first.
• Jump back into Databricks: Every completed answer includes a deep link to the full conversation in the Databricks UI, so you can dig in further.

Because Go uses the context of your active window and connected sources, you can also click Add sources to chat to point it at Databricks before asking your question. Go acts using the service principal’s Databricks permissions—the same for everyone using this configuration—so it can only access the data your organization has chosen to share through the connector.