Databricks
×
Papercrane
Everything lands in Databricks. Answers still leave as screenshots.
Papercrane connects to your SQL warehouse, browses Unity Catalog, and builds hosted dashboards from the tables you already trust. Then it joins them live with Salesforce, Stripe, GA4, and QuickBooks, and the people who open the result never need a Databricks login.
Connect in about a minute with a token you create in your workspace settings. Free to connect and ask. Paid plans start when you publish a dashboard. See pricing
The lakehouse is upstream of everyone. You are the last mile.
The Slack DM
"Can you pull renewals against usage real quick?" The usage gold table exists. The join to the CRM does not, and quick is never quick.
The export
The number leaves as a CSV, becomes a chart in a deck, and dies there. Next quarter someone screenshots the stale version.
The BI layer
Getting it into Tableau or Power BI means a semantic model, a license conversation, and one more copy of the logic to keep in sync.
Two ways to answer a question the catalog can't
The question of the day always spans one tool that never got ingested. Say it's "Which accounts renewing next quarter have falling product usage?"
Through the lakehouse
The route for data not yet in Unity Catalog
Stand up a CRM connector
you
Land bronze
you
Model silver
you
Publish the gold table
you
Add it to a Genie space
you
Write instructions, benchmark the answers
you
The business asks the question
finally, them
With Papercrane
The same question, resolved live
Connect Databricks and Salesforce
Ask the question
The agent plans the join, queries your warehouse and the CRM live
Builds a hosted dashboard
When data belongs in the lakehouse long term, build the pipeline; that instinct is correct. Papercrane is for the questions that can't wait for one, and for the joins that never justify one.
The "build a lightweight portal" ticket
Ask how to put lakehouse data in front of customers and the standard answer is an app around embedded dashboards, with your own tenant mapping and auth. That is the layer Papercrane hosts: each customer gets an isolated view of their own data, embedded in your product or on a custom domain, and nobody gets registered in your Databricks account.
What actually runs against your workspace
Connect
A token you create in your workspace settings. It carries the permissions of the identity that creates it, so for a scoped connection, create a service principal with only the grants you want Papercrane to have and mint the token as that principal. Revoke it in Databricks any time.
Plan
The agent browses Unity Catalog (catalog, schema, table) to find the right tables before it writes a line of SQL.
Query
Read queries on your Delta tables, through the SQL warehouse you point it at. Nothing installed in your workspace, no jobs scheduled, no compute beyond that warehouse. Connect several warehouses and address each by name.
Serve
Dashboards are hosted by Papercrane and refresh on the schedule you set, so your warehouse wakes for refreshes, not for every viewer. When a refresh fails, the error routes back to the agent, which repairs the dashboard.
Every dashboard is code you own. No lock in.
Questions that cross the catalog line
Gross margin by customer: usage from the lakehouse, invoices from Stripe.
Which channels bring signups that still retain at day 90?
Ticket volume per account against seat count and contract value.
What did activation look like the week after the release, by plan?
Every answer comes back as a chart on a hosted dashboard that refreshes on schedule, not a paragraph you paste into Slack.
Three ways off the last mile
| The job | AI/BI Dashboards + Genie | Tableau or Power BI on top | Databricks + Papercrane |
|---|---|---|---|
| Answer a question that spans a tool outside Unity Catalog | Ingest it first: connector, bronze, silver, gold, then curate | Same pipeline first, then model it again in the BI layer | The agent joins your warehouse with the live tool in one conversation join |
| Put a dashboard in front of your team | Strong, and Databricks One makes viewing free for account users | Licensed seats per viewer | A link to a hosted dashboard |
| Put a dashboard in front of customers or clients | Register them in your account, or build a portal around embeds | Per viewer licensing gets expensive fast | Isolated per customer views, embedded or on a custom domain |
| Keep the answer alive | You curate the Genie space and benchmark it | Workbooks and extracts to maintain | Errors route back to the agent, which repairs the dashboard |
| Warehouse behavior | Interactive viewers query the warehouse | Live connections or extract refreshes | Refreshes on your schedule; viewers open a hosted page |
When you don't need us
If your audience is registered in your Databricks account and the data they need is already modeled in Unity Catalog, use AI/BI Dashboards and Genie: they ship with the platform, cost nothing extra, sit closest to the data, and Databricks One now gives business users a genuinely clean way in. Papercrane is the right call when the answer needs tools that never made it into the catalog, or an audience outside your account: customers, clients, or the exec who was always going to get a screenshot.
Questions Databricks users ask us
The best Databricks questions join what never got ingested. Most teams connect Salesforce for the pipeline, Stripe for revenue, and GA4 for the front of the funnel.
Connect your workspace, ask the question that spans the tools around it, and hand back a live dashboard instead of a screenshot.
Free to connect and ask. Paid plans start when you publish. See pricing
Papercrane
Integrations: