Direct answers, including the awkward ones.

Yes. The Enterprise License is the same software. It runs as one self-contained container, on a Docker host your own team owns. Both editions carry the whole product.

One container image. Inside it, a gateway supervises every governed tool and brokers one shared Tableau session. The tools answer on loopback, not on your network. There is one thing to run and one thing to keep current.

It installs nothing there: no agent, no plugin. It signs in with a credential you issue, then works through the interfaces Tableau already publishes. It can never reach further than the person whose credential it holds.

Both. Their content models are close enough that the same governed tools work against either one. Where they differ is exactly what a migration review has to list.

English, Japanese, German, French, Italian and Russian, across every tool and the admin console. This website speaks the same six. Pick one here and you meet the same choice when you sign in.

There are 14 governed tools behind one shared sign-in, opened with your own Tableau token. You sign in once, then move between them freely, with no second sign-in.

A username, a password, and a code from an authenticator tool. The session is issued after both factors, and only after both. Every setup of this product works that way.

Yes. Any SAML 2.0 identity provider connects, one per client email domain. Home-realm discovery sends each person to the provider that owns their domain. Single sign-on ships dormant: the button and the route appear once an admin connects a provider.

Each one is hashed with scrypt at N=2^17. It is slow and memory hungry by design. A stolen file is worth little to whoever took it.

A wrong email and a wrong password do the same work and take the same time, so timing reveals nothing about who is on the roster. Rate limits and lockout run per account and per address, and both are checked before the costly hashing.

Every live session they hold ends the next time it is used. It does not wait for the token to expire. Roles are read fresh from the directory on every request, and they never travel inside a token.

Two timers run at once: thirty minutes idle, and twelve hours absolute. Activity slides the idle timer. The twelve hours do not move.

Your own personal access token. It carries exactly your permissions. The product sees what you see and nothing more. Revoke the token in Tableau and its reach ends at once.

Where a job has to run with nobody present. A scheduled pull signs in with the stored personal access token first, then falls back to a connected-tool JWT. The run records which credential answered.

Encrypted at rest, with a key derived for that one purpose. Files are written atomically at tight permissions. Once a secret is saved, no browser ever sees it again.

It is optional. Every tool produces its full result on its own deterministic engine, and AI is a box a person ticks for a single run. Leave it off and nothing reaches a model provider at all, wherever you run the container.

Yours. Boreon supplies no key and holds no shared pool. Your own contract with your provider is the one that applies. A key can be set for one person, or for the whole site.

Anthropic, OpenAI, Azure OpenAI, or any OpenAI-compatible endpoint. That last one matters on a strict network: point it at a gateway your platform team already runs, and a prompt travels only to a machine you operate.

Every change it suggests reaches a person first. It arrives as an approve-to-run card, and that card enters the same gate the owning tool already has, under the same signature and dual-control rules your governance team wrote.

When a question needs the data, it runs real SQL against a per-user copy of the table, not reasoning over a sample of it. The query engine is local and cannot reach the file system. One person cannot name another person’s copy.

By default, the words are not kept. What the admin console records is the shape of the usage: calls, tokens in and out, which model answered, and a cost. The cost is labelled an estimate.

Yes. TabTotal publishes a Model Context Protocol endpoint with read-only tools. It signs in with your own Tableau token. It ships switched off, and an admin turns it on when your firm is ready.

A governed Tableau workbook. Every figure in it comes from a live query on your own session, never invented to look plausible. Publishing takes an approval plus a re-typed signature, and it always makes a new workbook. What you already have is still there in the morning.

Neither is cut down: the trial runs the complete product, with the same tools, semantic layer, AI and audit trail. They are not two things to choose between, though. The Enterprise License is the product, and the published rate is its rate. Boreon’s TabTotal Cloud is a one-user, one-month trial on an instance we host, and it is where most Enterprise evaluations start.

The Enterprise License rate is published on the Editions page, priced by your total Tableau Creator Licenses, annually or on a three-year term. Deployment detail and service terms are agreed with you directly.

One user, one month, on us. Point it at a real Tableau site, not a demo dataset.

You do. Your team holds the bootstrap Super Admin account and the secret that keys every stored credential. Boreon holds neither.

Yes. All 14 governed tools, the interface languages, one semantic layer over Salesforce Data 360, Snowflake, Databricks and the Palantir AIP Ontology, and the same audit trail. These belong to the software itself.

That is a conversation to have with us. Both editions run the same container. The rest depends on what you are moving.

It inherits them and reports on them. Your grants stay the boundary of what anyone can reach, and the product holds no second opinion about who may see what.

What you get is mapped audit evidence. Guard tags every finding with a CIS Controls v8.1 Safeguard ID and a NIST CSF 2.0 function, and that tag travels with the finding into its CSV, its SARIF, its PDF and the screen. This product issues no certification.

The semantic layer stores meaning rather than rows: entities, dimensions, measures, and the tables they point at. Where the assistant needs data to answer a question, it builds a bounded, per-user copy for that job. The row count is what gets recorded, never the values.

It reads. The credential you grant is read-only, and the layer reaches exactly as far as that credential. You can revoke it in your own warehouse without asking us.

Two. That is where the code currently stops, not a design decision.

It inherits it. Tables, columns, comments, tags, keys and grants arrive with the first sync, and they keep their own names and wording on the way through. Nothing here asks your team to type any of it a second time.

At every point where something would change. The software may find things, explain them and draft the work. A named person approves, and that approval goes into the record your auditor reads next year.