tool

AI SDK

An AI SDK is the official client library a vendor ships for calling their model APIs — Anthropic SDK, OpenAI SDK, Google GenAI SDK, Mistral SDK, Vercel AI SDK (multi-vendor wrapper). Handles auth, retries, streaming, types.

Calling HTTP APIs directly works but loses ergonomics: each vendor's quirks (auth header naming, streaming format, error retry policy) become app code. AI SDKs abstract this: official client libraries handle auth, automatic retries with backoff, streaming responses, type safety, tool use round-trips. 2026 landscape: vendor-official SDKs (Anthropic SDK, OpenAI SDK, Google GenAI SDK, Mistral SDK, Cohere SDK), multi-vendor wrappers (Vercel AI SDK, LiteLLM, Vellum), language-specific (TypeScript / Python first-class, Go / Rust / Java second). Production tips: pin SDK version (breaking changes happen), use multi-vendor wrappers if you need portability, monitor SDK changelog for new features (prompt caching, computer use, MCP arrived through SDK updates).

When to use ai sdk

Any production LLM integration.

Common mistakes

Hand-rolling HTTP calls — miss retries, streaming, type safety.
Pinning to ancient SDK — miss new features for months.

FAQ

What is ai sdk?

When should I use ai sdk?

Any production LLM integration.

What are the most common mistakes with ai sdk?

Hand-rolling HTTP calls — miss retries, streaming, type safety. Pinning to ancient SDK — miss new features for months.

LLM gateway — An LLM gateway is the proxy layer between your app and one-or-many LLM providers — handles routing, fallback, caching, cost tracking, rate limiting, and observability. OpenRouter, LiteLLM, Portkey, Helicone, Cloudflare AI Gateway are 2026 leaders.
Model name portability — Model name portability is the principle of writing app code that survives swapping the underlying LLM provider (Claude → GPT → Gemini → local) without rewriting prompts, tool definitions, or response parsing.

Last updated: 2026-06-01. Raw markdown: https://promtable.com/glossary/ai-sdk.md.