Vanie AI logo
HelpWorkspaces & billing

Bring your own Gemini key: a full BYO setup guide

Most Vanie AIteams cut their monthly bill by 75 to 95 percent the day they switch to BYO. This guide walks you through the entire setup in plain English: what BYO means, how to get a Gemini API key, how to wire it into a workspace, and what to expect when Google's invoice arrives. No coding required.

What BYO actually means

BYO stands for Bring Your Own. In Vanie AI, a BYO workspace is one where you connect your own Google Gemini API key instead of using the pre-paid credits that Vanie AI sells. When the workspace makes a voice request, the bytes flow throughVanie AI but the model call is billed against your Google account.

The difference matters because Vanie AI prices Platform credits with margin baked in. We have to: that margin covers our infrastructure, support, model-routing logic, and the eng team that keeps everything running. Plenty of customers are happy with that trade-off and stay on Platform forever. But heavier users, and especially teams already on a Google Cloud contract, almost always come out ahead by paying Google directly and treatingVanie AI as a flat subscription on top.

The savings vary by workload, but for live voice agents BYO typically trims the per-minute credit rate by roughly 75 to 95 percent, because you skip our per-token markup entirely. Your mileage will depend on your Google tier and how much you negotiate with their sales team, but everyone we have moved across has been net cheaper within the first billing cycle.

Before you start: what you need

The setup itself takes about ten minutes. The prerequisites are the part that trips people up, so check this list before you open the dashboard:

  • A Google account. Personal Gmail works for experimentation. For production, use a Google Workspace account tied to your company so the billing trail lives somewhere your finance team can see.
  • A Google Cloud project with billing enabled. Without billing, the Gemini API stays on a tiny free tier that will run out within minutes of serious traffic. Adding a card unlocks the paid tiers.
  • The Gemini API enabled in that project.Search for “Generative Language API” in the Google Cloud console and click Enable. This is a one-click operation but it is mandatory.
  • An API key from Google AI Studio. This is the actual secret you will paste into Vanie AI. We cover the exact URL below.
  • Owner or Admin permission on the Vanie AI organisation. Members can use a BYO workspace but only Admins can create one.

If you do not yet have a Google Cloud project, do not worry: AI Studio will create one for you automatically the first time you generate a key. It is free to create the project. You only pay for what you use.

Step-by-step: connecting your key

Open the Vanie AI dashboard and follow along. Each step corresponds to one screen.

1. Open the workspace switcher

At the very top of the dashboard, next to the Vanie AI logo, you will see your current workspace name with a small chevron next to it. Click it. A panel slides down listing every workspace you belong to.

2. Click + Create workspace

At the bottom of that panel is a button labelled + Create workspace. Click it. A modal opens asking for a name and a type.

3. Choose a name and the BYO type

The name is for you and your team. We recommend something descriptive like “Acme Production BYO” or “Marketing Sandbox” so it is obvious at a glance which key is being used. Names can be edited later.

For the type, you will see two cards: Platform (pay Vanie AI in credits) and BYO (bring your own provider key). Click the BYO card. The modal expands to show a provider section.

4. Pick Gemini, paste, and test

At the moment Gemini is the only supported BYO provider, which is intentional. We support exactly one provider so we can keep the runtime behaviour identical across all your features. Pick Gemini from the dropdown.

Paste your key into the API key field. You will see a green key icon if the format looks right. Click Test.Vanie AI sends a tiny ping to the Gemini API to make sure the key is valid, the API is enabled, and the key has quota available. A green check appears within a couple of seconds if everything is healthy.

5. If the test fails: common fixes

Three errors come up about ninety percent of the time. Here is how to fix each:

  • “API key not valid”. You probably copied the key with a trailing space or a hidden character. Go back to AI Studio, click the key, and use the copy button inside the panel rather than triple-clicking the text.
  • “Generative Language API has not been used in project”. The Gemini API is not enabled on the Google Cloud project that owns this key. Open the console, enable the API, wait sixty seconds, then click Test again.
  • “Billing account required”. The project is still on the free tier and the key has been throttled. Open the Billing tab in Google Cloud, attach a card, and retry.

If you get an error that does not match any of the above, click the small View details link next to the red banner. It will paste the raw Google error message into a card so you can share it with support if needed.

6. Save and switch in

Once the test goes green, click Save. A toast appears at the bottom right reading “Workspace created · switched in”, and Vanie AI automatically reloads the dashboard inside the new workspace. You can confirm by looking at the top bar: the workspace name now reads whatever you chose, and a small BYO pill sits next to it.

That is it. Every voice request you make in this workspace from now on will be billed against your Google account, not yourVanie AI credit balance.

Where to find your Gemini API key

The canonical URL is aistudio.google.com/app/apikey. Sign in with the Google account you want to bill. If you have never visited AI Studio before, you will be asked to accept the terms of service and pick a country.

The page shows a table of any keys you already have, plus a prominent blue button labelled Create API key. Click it. A dropdown appears listing your Google Cloud projects; choose the project you want to charge or let AI Studio create a fresh one. After a second or two the new key appears in a modal with a copy button. Use that copy button: the value is only shown once in full, and on subsequent visits AI Studio masks the middle of the string.

Treat the key like a password. Do not email it, do not paste it into a shared chat, and do not commit it to git. InsideVanie AI it is encrypted; outside Vanie AI you should assume that anyone who reads it can spend money on your account.

Quota tiers, in plain English

Google rate-limits the Gemini API in three tiers. Knowing which one you are on is the difference between a smooth rollout and a wave of failed calls on launch day.

  • Free tier. The default for any new project. Generous enough to test a demo, far too small for production: think tens of requests per minute, not thousands. Vanie AIwill work, but anything beyond a single user hitting it interactively will throttle within seconds.
  • Tier 1. Unlocked the moment you attach a billing card. Limits jump roughly a hundredfold. Most early production teams sit here happily for the first few months.
  • Tier 2 and above. Unlocked automatically after you have spent a certain amount on Gemini (the exact number changes; Google publishes it on the rate-limits page). Limits jump again, and at this point quotas stop being the bottleneck for normal workloads.

To see your current tier and live usage, open the Google Cloud console, navigate to APIs & Services · Quotas & System Limits, and filter by “Generative Language API”. The same view shows requests per minute, tokens per minute, and the headroom you have on each. Vanie AI surfaces a simplified copy of the same numbers under Workspace · Usage · Quota so you do not have to switch tabs.

What to do when your Google bill arrives

Google bills monthly, on the same date each month, against the card or invoice account attached to the project. The amount is the sum of every token, second of audio, and second of live session you used. Vanie AI does not see a single rupee of that transaction: it goes straight from your bank to Google.

When the invoice lands, the line items will look like Generative Language API · gemini-2.5-flash · input tokens. To reconcile this against what you used inside Vanie AI:

  • Open Workspace · Usage in Vanie AIand switch the date range to the same calendar month.
  • The totals at the top — input tokens, output tokens, audio seconds, live seconds — match the Google line items one to one. Multiply by Google's public per-unit rate and you should land within a percent or two of the invoice.
  • If the numbers diverge meaningfully, check whether anyone on the team also used the key outside Vanie AI. Shared keys see shared spend.

For tax handling, the invoice comes from Google's entity in your country (Google Cloud India for Indian customers, Google Ireland for EU, and so on). Your finance team can claim it the same way they handle any other Google Cloud expense.

Switching back to Platform later

Nothing about going BYO is irreversible. If you want to flip back, you have two options.

Quick toggle. Open the BYO workspace, go to Settings · Billing mode, and click Switch to Platform. The workspace immediately starts drawing from your organisation's credit pool instead. Your Gemini key stays stored (still encrypted, still private) so you can switch back without re-pasting it.

New workspace. If you want to keep both modes around — say, BYO for production and Platform for experimentation — just create a second Platform workspace from the switcher. You can have as many of each as you like. The top bar always shows the active mode so you know which one you are spending in.

Frequently asked questions

The same five questions come up every onboarding call. Skim these first before you reach out to support.

How much do I save by bringing my own Gemini key?

Most teams save between 75 and 95 percent of their previous Vanie AI credit spend, depending on volume. You pay Google directly at standard Gemini API rates, and Vanie AI only charges a flat workspace subscription instead of marking up tokens.

Does Vanie AI see or store my API key?

Your key is encrypted at rest with per-workspace envelope encryption and is only decrypted in memory at request time. It never appears in logs, support tickets, or analytics. You can rotate or revoke it any time from Workspace settings.

What happens if I hit the Gemini quota?

Calls fail closed with a clear quota-exceeded message. Vanie AI does not silently fall back to a Platform credit pool. Raise your Gemini tier in Google Cloud, or temporarily switch the workspace back to Platform mode while you wait.

Can I share one Gemini key across multiple workspaces?

Yes. Paste the same key into each workspace. Just remember that all of those workspaces share a single Google quota and a single Google bill, which can make it harder to track per-team usage.

Do BYO workspaces support every Vanie AI feature?

Almost. Anything that uses Gemini for speech to text, text to speech, voice agents, or live conversations works identically. A small number of premium features built on Vanie AI-managed models are Platform only and clearly labelled as such in the dashboard.

Next steps

Once your BYO workspace is live, the next thing worth doing is setting a budget alert in Google Cloud so you know the moment spend crosses whatever threshold matters to you. Then point a real workload at it: spin up a voice agent and place a few test calls. The credit counter in the top bar will stay flat; the savings will show up in your Google invoice at the end of the month.

If anything in this guide felt unclear, or your Test button is still red after working through the common fixes, ping support from the question mark menu inside the dashboard. We answer in business hours and we are happy to share a screen.