Vanie AI workspaces, explained from scratch

What is a workspace?

A workspace is the unit of isolation in Vanie AI. Everything that costs money, holds state, or makes a model call lives inside exactly one workspace. That includes your subscription plan, your credit balance, your voice agents, your knowledge bases, your call recordings, your transcripts, your usage charts, your API keys, and your AI provider configuration.

Two workspaces under the same login are as separate as two entirely different companies. They cannot see each other's recordings. They cannot share credits. A change to the model routing in one has zero effect on the other. The only thing they have in common is you: your user identity is a thin layer that sits on top and decides which workspaces you are allowed to open.

This matters because most things you might expect to be account-level — your plan, your AI keys, your agents — are actually workspace-level. If a teammate says “we're out of credits”, the right question is always which workspace. The top bar always shows the answer.

Why would I want more than one?

You start with one workspace and a lot of teams happily stay there forever. The teams that branch out tend to fit one of three patterns.

Separate environments

Engineering teams almost always want a dev, a staging, and a prod workspace. Different keys, different agents, different call recordings, different spend dashboards. A prompt change in dev cannot accidentally talk to a real customer. A spike in staging cannot drain the prod credit pool. The blast radius of any change is contained to one workspace.

Separate clients in an agency setting

Agencies running voice agents for multiple end customers use one workspace per client. Each client gets their own plan, their own monthly invoice, their own recordings retained for the agreed period, and their own provider setup. When a client offboards, the agency soft-deletes that one workspace and everything else stays untouched. Pricing per client is the workspace's subscription, full stop.

Experimenting with BYO without losing Platform

Plenty of teams want to test bringing their own Gemini key but are nervous about touching their working Platform setup. The answer is a second workspace. Spin up a BYO workspace alongside the original Platform one, point a small slice of traffic at it, compare cost and latency, and either keep both or migrate wholesale. Your original Platform workspace, with all its agents and recordings, is never at risk during the experiment.

The default workspace you get at signup

When you sign up, Vanie AI creates one workspace for you automatically and drops you straight into it. That default workspace is a Platform workspace. Platform means Vanie AI pays the underlying AI bill: every voice request you make is charged against the credit balance in your workspace, and behind the scenes we pay Google for the Gemini tokens.

The default workspace is named after your organisation but you can rename it in Settings · General at any time. It comes with a small trial credit balance so you can try the product before you put a card down. Once you do, the plan and credit balance live on this workspace and only on this workspace — adding a card here does not auto-fund any second workspace you create later.

Creating a second workspace

From the workspace switcher in the top bar, click + Create workspace. The modal asks for a name and a type.

The nameis whatever helps you and your team recognise it: “Acme Production”, “Marketing Sandbox”, “BYO Experiment”. Names can be edited later, so do not overthink the first pass.

The type is the important choice. You pick either:

• Platform — Vanie AI handles the AI provider relationship, you buy credits from us, we pay Google. Fastest to set up, zero key management.
• BYO — Bring Your Own. You paste a Google Gemini API key, we use that key on every request in this workspace, Google bills you directly. Cheaper at volume but you own the Google quota and the Google invoice. See the BYO setup guide for the full walkthrough.

Once you save, Vanie AI creates the workspace, attaches a fresh empty credit balance (Platform) or a fresh provider config (BYO), and switches you in.

Switching between workspaces

The topbar dropdown is the only way to switch. Click your current workspace name, see the list of workspaces you belong to, click the one you want, and Vanie AI reloads the dashboard inside the target workspace. No re-login required.

Under the hood, switching swaps your session JWT. Vanie AIuses a 28-day JWT for web sessions, encoded with the active workspace id. When you pick a new workspace from the dropdown, the backend issues a fresh 28-day JWT bound to that workspace, writes it back into the session cookie, and the page reloads. The previous JWT is invalidated server-side immediately. There are no refresh tokens to manage and no double-logins; the swap is atomic and instant.

One small thing to know: because the JWT is bound to a single workspace at a time, opening two browser tabs and switching workspaces in one of them does not retroactively change the other tab. We cover what that means for an active voice call below.

Per-workspace billing

Every workspace has its own billing stack. Concretely:

• Its own plan. One workspace might be on the Starter plan, another on Scale, a third still on the free trial. Plans are picked per workspace in Settings · Billing.
• Its own balance. Credits, top-ups, and rollover live on the workspace, not on the user. Topping up workspace A does not change the counter in workspace B by a single credit.
• Its own subscription. A separate Stripe subscription, a separate monthly invoice, a separate VAT line. If you change cards in one workspace, the others keep using their existing cards.
• Its own provider setup. Platform mode versus BYO mode, model routing preferences, and provider quotas all live at the workspace level.

The intentional consequence is that workspaces never share credits. If finance wants to see “total Vanie AIspend” across your org, the right view is the organisation-level billing report which sums invoices across workspaces. There is no shared pool that workspaces draw from by default.

Per-user workspace limit

A Vanie AI user can belong to up to three workspaces by default. This is a deliberately low ceiling: in practice it covers the dev/staging/prod case for most engineers and the two-or-three-clients case for solo agency operators.

Organisation admins can raise the limit for individual users or for the whole org through support. Agencies running a workspace per client routinely have admins on counts in the dozens; that is fine, it just needs an explicit bump because most users do not need it.

The limit is per user, not per organisation. An organisation can own many more workspaces than three; the cap only governs how many a single user can be a member of at the same time. You can leave a workspace at any time from Settings · Members to free up a slot.

What if I switch workspaces mid-call?

This question comes up the first time an engineer opens two tabs and starts experimenting. The short answer: nothing breaks. The longer answer is worth knowing.

When you start a voice call, our gateway opens a server-side session bound to the workspace that was active at the moment of connect. That session holds the model routing, the credit source, and the recording destination for the entire call. It is independent of your browser cookies from that point on.

If you switch workspaces in another tab while the call is live:

• The active call continues against its original workspace until it ends naturally. Audio keeps flowing, the transcript keeps writing to the original workspace, and credits are deducted from the original workspace. The user on the other end sees nothing.
• Any new action you take after the switch — starting a new call, deploying an agent, topping up credits — hits the new workspace. The dashboard you are looking at reflects the new workspace immediately.
• When the live call eventually ends, the recording and transcript land in the original workspace. They are not migrated, even if you have switched.

The net effect is that there is no race condition you can accidentally trigger by switching. We picked this model specifically so support engineers can hop into a customer workspace to debug without disrupting their own live agents.

Deleting a workspace

Deletion is two-stage by design, because nothing in voice AI is more painful than restoring data you thought you had.

Stage one: soft delete. From Settings · Danger zone, an admin clicks Delete workspaceand types the workspace name to confirm. The workspace immediately disappears from the switcher, all open sessions for that workspace are revoked, the subscription stops billing, and the workspace enters a 30-day grace period. During those 30 days an admin can restore the workspace with one click from the organisation's archived workspaces view. Everything comes back exactly as it was: agents, recordings, transcripts, members, balance.

Stage two: hard delete. After 30 days, the workspace is hard deleted. This is irreversible. Vanie AIpermanently removes call recordings, transcripts, agent configurations, knowledge bases, custom prompts, and any stored API keys. We retain only the minimum billing metadata required for our legal and tax obligations, with personal data scrubbed.

If you need to delete sooner than the 30-day window (for a DPDP / data-subject erasure request, for instance), contact support and we can fast-track the hard delete with the appropriate paperwork.

Workspace types: Platform vs BYO

The two workspace types behave identically from the user interface up. Voice agents work the same. The knowledge base works the same. The dashboard looks the same. What changes is who pays the AI provider, and that decision ripples into pricing, quotas, and key management.

Pick Platform if you want zero key management and predictable prepaid credits. Pick BYO if you have meaningful volume, an existing Google Cloud relationship, or finance constraints that prefer one consolidated Google invoice. The full BYO walkthrough lives in the BYO setup guide.

Side by side

Frequently asked questions

These are the questions support hears most often during the first week with a new workspace.

Next steps

If you are still on the default Platform workspace and starting to feel the credit spend, the natural next step is to follow the BYO setup guide and stand up a second workspace pointed at your own Gemini key. If you are setting up environments for an engineering team, create the dev, staging, and prod workspaces up front and add only the people who should see each — Members can be scoped per workspace from Settings · Members. If you are an agency, name workspaces after clients and use the per-workspace invoice as a clean billing artefact.

Anything we have not covered here, the question mark menu inside the dashboard opens a chat with the same engineers who built this. We answer in business hours.