Skip to content

Bring your own API key

Create an OpenAI, Anthropic, Gemini, or OpenRouter API key and connect it to the Verxion native agent.

The Verxion native agent uses BYOK: bring your own API key. You choose the AI provider, create a key in that provider’s console, and paste it into Verxion on your device.

Verxion does not host the model for you and does not store the key on our servers. The key is stored encrypted on your device and sent only when the native agent needs to talk to your selected provider.

Before you start

You need:

  1. A provider account: OpenAI, Anthropic, Google, or OpenRouter.
  2. API access enabled for that account.
  3. Billing or free credit available on the provider side.
  4. The Verxion native app installed and signed in.

A ChatGPT Plus, Claude Pro, or Gemini Advanced subscription is not the same thing as API credit. The native agent uses provider APIs, so usage is billed by the provider account that owns the key.

Treat the key like a password. Copy it once, paste it into Verxion, and do not share it in chat, email, screenshots, or support messages.

OpenAI

Use OpenAI if you want GPT models in the native agent.

  1. Open OpenAI Platform API keys.
  2. Sign in with the OpenAI account that should pay for API usage.
  3. Create a new project key.
  4. Copy the key when OpenAI shows it.
  5. In Verxion Native, open the agent setup screen, choose OpenAI, paste the key, choose a model, and validate.

OpenAI keys usually start with sk-proj-. OpenAI only shows the full secret when you create it. If you lose it, create a new key and remove the old one. See OpenAI’s help page: Where do I find my OpenAI API key?.

Anthropic / Claude

Use Anthropic if you want Claude models in the native agent.

  1. Open Anthropic Console API keys.
  2. Sign in with the Anthropic account that should pay for API usage.
  3. Create an API key.
  4. Copy the key.
  5. In Verxion Native, open the agent setup screen, choose Claude, paste the key, choose a model, and validate.

Anthropic keys usually start with sk-ant-. Anthropic’s getting-started guide also covers the API key requirement: Get started with Claude.

Gemini

Use Gemini if you want Google’s Gemini models in the native agent.

  1. Open Google AI Studio API keys.
  2. Sign in with the Google account that should own the key.
  3. Create a new Gemini API key.
  4. Copy the key.
  5. In Verxion Native, open the agent setup screen, choose Gemini, paste the key, choose a model, and validate.

Gemini keys usually start with AIza. Google’s guide explains key creation and key types here: Using Gemini API keys.

OpenRouter

Use OpenRouter if you want to choose among many models with a single key — including budget options (DeepSeek, Qwen, Gemini Flash) and frontier models. OpenRouter is a gateway: it routes your request to the model provider you pick, under your own OpenRouter account.

  1. Open OpenRouter API keys.
  2. Sign in with the OpenRouter account that should pay for usage.
  3. Create a new API key (optionally set a credit limit on it).
  4. Copy the key.
  5. In Verxion Native, open the agent setup screen, choose OpenRouter, paste the key, choose a model, and validate.

OpenRouter keys usually start with sk-or-v1-. Verxion’s picker shows a curated list of supported models, with the most reliable ones marked as recommended. For your privacy, Verxion forces no-retention and no-training routing on every request (zdr + data_collection: deny), regardless of your account settings. Create and manage keys here: OpenRouter API keys.

Connect it in Verxion Native

  1. Open the Agent tab.
  2. If no key is connected, tap your provider.
  3. Paste the API key.
  4. Pick the model you want to use.
  5. Tap Validate and connect.

Verxion checks the key before saving it. If validation passes, the agent is ready. If validation fails, the app shows a safe error category such as invalid key, out of credit, rate limited, provider unavailable, or timeout.

Billing and limits

Your provider bills your API usage directly. Verxion does not charge for model usage and does not control the provider invoice.

Recommended: set a budget before you connect the key. Provider consoles let you configure spend limits, monthly budgets, or budget alerts depending on the provider. This is one of the best ways to control what the native agent can spend while you test models, compare providers, or leave the agent available on your device.

Set real spend limits in the provider console:

Verxion may show estimated cost information for agent turns, but the provider dashboard is the source of truth.

Security and revocation

Your key is stored encrypted on this device. Verxion sends it per request to the selected provider through the agent harness and never stores it server-side.

If you think a key leaked:

  1. Revoke or delete it in the provider console.
  2. Create a new key.
  3. Open Verxion Native and replace the old key.

Removing the key from Verxion stops this device from using it, but it does not revoke the key at the provider. Revoke leaked keys in the provider console.

Troubleshooting

The key is invalid

Check that you pasted the full key, not a label or project id. If the provider only showed it once and you did not save it, create a new one.

The provider says you are out of credit

Add billing or credits in the provider console. Verxion cannot add provider credit for you.

Requests are rate limited

Wait a moment or choose a provider/model with higher limits. Rate limits are controlled by the provider account.

The provider is unavailable

Try again later or temporarily connect another provider.

I changed devices

Add the key again on the new device. BYOK keys are stored locally per device.