Aiolum Browse stories
All tools
Building the product · 2 of 252 stories

Claude Agent SDK

A software library from Anthropic for building AI agents in Python or TypeScript: the Claude model plans steps and calls tools such as file editing, shell commands and web search.

What Claude Agent SDK is

Written from its maker's own pages, read on 28 October 2026.

The Claude Agent SDK is a library from Anthropic for building AI agents. An agent is a program that completes a task by planning its own steps and calling tools that read files, run commands or edit code. The SDK gives a Python or TypeScript program the same tools, agent loop and context management that run Claude Code, Anthropic's coding agent; in fact the package bundles the Claude Code command-line program and drives it.

A developer sends a prompt and options, and the SDK returns a stream of messages while Claude works. Built-in tools read, write and edit files, run shell commands and search the web. Developers add their own tools as functions or through MCP (Model Context Protocol, a standard way to connect an AI model to outside tools and data). Permissions decide which tools run without approval, hooks run custom code at fixed points of the loop, subagents take focused subtasks, and sessions keep context so that work can resume later.

The SDK is free to install; the Python repository carries the MIT license. Model use is paid through a Claude API key, billed per token.

Made by
Anthropic
Price
SDK free; Claude API usage is pay per token, for example Sonnet 5.5 at US$2 per million input tokens and US$10 per million output tokens (read 28 October 2026)
Free plan
SDK free; API usage paid

Who uses it, and for what

One row per story. The grade is for the claim that the business uses this tool, not for its revenue.

StoryWhat it does thereWhenCost statedGradeSource
PolsiaFramework on which the agents are built: the model calls tools such as a code editor, an inbox or an advertising account.CPolsia Terms of Service and app code, read 12 October 2026
RunbearAgent runtime on which the newer agents run, named in the changelog of October 2026.since 2026CRunbear changelog, first page, 13 October 2026 entry

Grade A: the tool's maker or a payment processor confirms the use. Grade B: an independent report, or the tool is visible in public code. Grade C: the founder or the company says so.

Used in the same businesses

Other tools in the stories above, with the number of stories that share both.

Other tools for the same job

Building the product, with the number of stories that use each.

Back to all tools