Tools & agents
SDK (for AI)SDK
Also known as: software development kit
A ready-made code library that makes it easier for developers to build with an AI model, wrapping the raw API in convenient tools.
An SDK is a toolkit for programmers that makes building with AI easier. Where an API is the raw doorway to a model, an SDK is a set of pre-built tools and shortcuts around that doorway, so developers write less code and make fewer mistakes. You never touch it directly, but it is a big part of why new AI apps can be built so quickly.
An SDK (software development kit) is a library of code, in a specific programming language, that wraps a provider's API with convenient functions, handling authentication, requests, retries, streaming and data types for you.
Why developers use it
- Less boilerplate: call the model with a simple function instead of hand-crafting network requests.
- Fewer errors: built-in handling of common cases.
- Higher-level features: many AI SDKs also provide helpers for tool use, agents and orchestration.
An SDK sits on top of an API: the API is the network interface; the SDK is the friendly code around it.
An SDK packages client libraries and abstractions over a model provider's API, managing auth, serialization, streaming, retries/backoff, typing and error handling, and increasingly higher-level constructs (agent loops, tool/function-call plumbing, structured-output helpers, orchestration primitives). It reduces integration effort and standardises idiomatic usage per language.
The layering, model → API → SDK → framework, matters for portability and lock-in: thin SDKs stay close to the API, while opinionated agent/orchestration frameworks add capability at the cost of coupling. Interoperability standards (MCP for tool/data connections, compatible request schemas) and provider-agnostic SDKs mitigate lock-in. Practical selection weighs language support, maintenance and community, feature scope vs simplicity, and how cleanly the SDK exposes underlying API controls (streaming, tools, versioning) needed for production systems.