AI Hub — API Access and Keys
What this hub covers
Section titled “What this hub covers”Before you can call a model API, you need a credential. Each provider manages keys differently: some require billing before they issue a key, some offer a free tier, and some (like Google AI Studio) allow free usage up to a generous quota without a payment method. This hub groups the setup guides by provider.
Suggested reading order
Section titled “Suggested reading order”-
API Tokens overview — read this first. Covers the security principles that apply to every provider: environment variable storage, rotation policy, least-privilege scoping, and what not to do (no hardcoding, no committing to source control).
-
Choose the guide for your provider:
| Provider | Access method | Free tier | Guide |
|---|---|---|---|
| OpenAI | Platform API key | $5 trial credits | OpenAI API Key |
| Google Gemini | AI Studio key | Yes (generous) | Google Gemini API Key |
| Anthropic Claude | Console API key | No (billing required) | Anthropic Claude API Key |
| Hugging Face | Access token | Yes (Inference API) | Hugging Face API Token |
| Cohere | Dashboard key | Yes (trial key) | Cohere API Key |
| Mistral AI | Console API key | No (billing required) | Mistral AI API Key |
Each guide follows the same structure: prerequisites, creating the key, storing it safely, and a minimal verification call.
Where to go next
Section titled “Where to go next”Once you have credentials:
- Build with the model directly: see the provider’s overview page (linked from Platforms and Products hub) for SDK examples and feature guides.
- Build a retrieval pipeline: see RAG vs agentic RAG for patterns that use the API alongside a vector store.
- Use an agentic tool: see the Coding Agents overview for tools like Claude Code and GitHub Copilot that consume the API on your behalf.