Kyb/DocsBeta
Start free

API Keys & Usage

Kyb uses AI for entity classification, cloud discovery abstraction, and the chat assistant. These features consume tokens. This page explains how token usage works and how to bring your own API key for unlimited access.

Understanding Token Usage

Tokens are a unit of AI processing. Every interaction with the AI assistant, every entity classification, and every cloud discovery abstraction consumes tokens. The number of tokens depends on the length and complexity of the conversation.

You can monitor your current token usage in the token meter, visible near the chat input in the Explorer. The meter shows how much of your monthly budget you have used.

Free Tier

Every Kyb account includes a free monthly token budget. This is enough to explore the product, build a small-to-medium architecture graph, and run cloud discovery on a typical GCP project. The budget resets monthly.

When you reach your free tier limit, AI features will be temporarily unavailable until the budget resets. You can still use all non-AI features (manual editing, views, export, team management) without limits.

Bring Your Own Key (BYOK): Multi-Provider

For unlimited AI usage and zero platform token consumption, you can bring your own API key. Kyb supports three leading frontier model providers:

Anthropic

Claude 3.5 Sonnet and Haiku models. Get your key from console.anthropic.com.

Google Gemini

Gemini 2.5 Pro and Flash models. Get your key from aistudio.google.com.

OpenAI

GPT-4o and reasoning models. Get your key from platform.openai.com.

Setting up BYOK

  1. 1. In Kyb, navigate to Settings > Billing.
  2. 2. Select your preferred AI Provider (Anthropic, Google, or OpenAI).
  3. 3. Paste your secret API key and click Save Key.
  4. 4. The active provider is highlighted with a green checkmark. All AI operations (classification, chat, cloud discovery abstraction) will now route directly through your key.

Security & Key Encryption

Your API keys are encrypted at rest using industry-standard AES-256-GCM encryption and used solely to execute model inference on your behalf. Keys are masked in the UI, never stored in client-accessible local storage, and can be removed or rotated at any time.

Reducing Token Usage

  • Batch descriptions -- Describe multiple systems in a single chat message rather than one at a time.
  • Deterministic cloud discovery -- Run discovery in deterministic-only mode to skip the AI abstraction step. This uses zero tokens but produces infrastructure-level entities instead of architecture-level ones.
  • Manual editing -- Use the entity form and drag-to-connect for straightforward additions that do not need AI classification.

Need help?

If you have questions or run into issues, reach out at [email protected]. You can also start over with the Getting Started guide.