Dashboards are rebuilt by hand
Every sheet is re-authored and every calculation rewritten in SQL, one workbook at a time.
Migrations and the LLM portal
The Boreon MCP Layer over Tableau, Snowflake and Databricks.
Move dashboards and Prep flows from Tableau Server or Tableau Cloud to Snowflake or Databricks, rebuilt on each warehouse’s own dashboard and data technology. Then the Boreon MCP Layer stays inside your firewall as the central chat portal for your data warehouses, yours to keep with no license fees. MVP in two weeks, go-live in one month.
- ~10% of one migration
- No consulting firm needed
Go live in one month
- Weeks 1 to 2MVPThe Boreon MCP Layer inside your firewall, with your priority dashboards moved and every number proven.
- Weeks 3 to 4Final deploymentThe rest of your priority workbooks, hardening, sign-off and training.
- Day 30Go liveYour team runs it in chat, and every future move is yours.
The problem
Why platform migrations stall
Most of the time goes on work that never makes it into the plan, and business users are still waiting for answers after the move.
Prep flows hide the real logic
The data a workbook reads is often shaped by a Prep flow nobody has documented.
Each platform needs its own experts
Tableau, Snowflake and Databricks each speak their own language, and business users wait on the specialists who speak it.
The migration path
From Tableau, to your warehouse
Every migration starts on Tableau Server or Tableau Cloud and lands on the warehouse’s own dashboard and data technology, from the same chat.
- To Snowflake. Dashboards become Streamlit in Snowflake apps, and Prep flows become Snowflake dynamic tables.
- To Databricks. Dashboards become Databricks AI/BI dashboards with Genie built in, and Prep flows become materialized views.
Every number is checked against Tableau’s own answer before anything is published on the warehouse.
Inside the Boreon MCP Layer
One semantic layer over all three
It reads each platform through its own APIs, keeps one model of how they map, and writes only when you ask it to. Every write is audited.
Agentic chat
Plain-words requests, live status and next-step chips, with answers that come back as live tables and charts.
Semantic layer
One model over all three platforms: Tableau fields mapped to Snowflake and Databricks columns, and kept current.
Migration engine
Workbooks, dashboards and Prep flows from Tableau Server or Tableau Cloud, planned and then proven on the warehouse.
Fidelity engine
Tableau’s own numbers compared with the warehouse’s, sheet by sheet, before anything is published.
Lineage
From Prep flow to data source to workbook, down to the column.
Governance
Read-only by default, every write audited, and your SSO roles carried through to both warehouses.
Proof
It publishes only what it can prove
Measured live on a real Tableau workbook moved to Databricks: DBX Platform Latency Calls & Uptime.
- All or nothing. If one sheet has no faithful match, nothing is published and the agent says why.
- Numbers first. Averages, distinct counts and medians stay exact under dashboard filters.
- Proven before it ships. Every sheet’s query runs on the warehouse before publishing.
- Read-only by default. The only writes are the ones you ask for, and each one is audited.
- sheets identical to Tableau
- 3 / 3
- fidelity score
- 96%
- rows in the view, built in 72 seconds
- 20M
- SQL errors on publish
- 0
More on the Boreon MCP Layer
Go deeper
Each part of the offering has its own page.
Tableau Server & Cloud
The Boreon MCP Layer over Tableau Server or Tableau Cloud on its own: ask in plain words, govern every workbook, on the LLM you choose.
Migration options
What moves where from Tableau Server or Tableau Cloud, how a migration runs, and the proof every move hands you.
Tableau to Snowflake
From Tableau Server or Tableau Cloud: dashboards become Streamlit in Snowflake apps, and Prep flows become dynamic tables.
Tableau to Databricks
From Tableau Server or Tableau Cloud: dashboards become AI/BI dashboards with Genie, measured on a real workbook.
LLM Portal
The chat portal that stays inside your firewall when the migration ends.
SDLC projects
How we deliver it: an MVP in two weeks, and go-live in one month.
Questions and answers
The things buyers ask first.
Which migrations does the Boreon MCP Layer run?
From Tableau Server or Tableau Cloud to Snowflake or Databricks. On Snowflake, dashboards become Streamlit in Snowflake apps and Prep flows become dynamic tables; on Databricks, dashboards become AI/BI dashboards and Prep flows become materialized views. Between the two warehouses it can also lift and shift tables, views and SQL.
How quickly can we go live?
An MVP lands in two weeks: the Boreon MCP Layer inside your firewall, your priority dashboards moved with their numbers proven, and the portal in your team’s hands. Final deployment and go-live follow within one month.
Where does it run?
As one central MCP server inside your own Azure VNet or AWS VPC, in a private subnet behind your firewall, beside your data.
Which AI does it use?
Your own OpenAI keys, through Azure OpenAI or OpenAI. The keys and the account stay yours.
What does it cost?
The project for the whole platform typically costs about 10% of a single data warehouse migration project, and there are no license fees. The platform stays with you in perpetuity; yearly maintenance, managed service and support plans are optional.
Do we need a consulting firm?
No. You integrate the Boreon MCP Layer once, inside your firewall, then migrate at your own pace in chat, with no deadline pressure and no bill per move.
Do our people keep their permissions?
Yes. Your SSO roles carry through to both warehouses, and people see what they are allowed to see.
What if a sheet has no faithful match?
Nothing is published. A move is all or nothing, and the agent names the sheet and the reason in plain words.
MVP in two weeks. Go live in one month.
Tell us which dashboards on your Tableau Server or Tableau Cloud site you would move first. We scope them, stand up the Boreon MCP Layer inside your firewall and put a working MVP in front of your team in two weeks, with final deployment and go-live inside the month.