Intel / repo

claude-agent-sdk-typescript v0.3.305

Source: github.com
Author: ashwin-antTop Viber
Date: Published Oct 20, 2026

Context

The TypeScript Claude Agent SDK 0.3.305 lets developers run Claude Code from their own programs, and this release fixes two places where the program's instructions were not fully honoured. The more important one concerns setMcpServers, which an application uses to hand the session its own servers. Previously, if the user's own settings contained a server with the same name, that server could replace the application's one later in the session. That way an embedded could end up talking to different tools than the developer configured, and the fix keeps the application's server in place. The same fix ships in Claude Code 2.1.305, which the release notes say the SDK is now at parity with.

The second fix is for startup(). When initializeTimeoutMs was passed as null, time spent in SessionStart or Setup hooks still counted against a 60-second limit. Null now behaves the same as leaving the option out. Relatedly, when startup() or prewarm() times out with no such running, the error now lists possible causes instead of steering people toward checking authentication and network. The release also adds recommendations and fix fields to the behaviour windows in the get_usage response, which appear to be the SDK counterpart of the one-line fixes Claude Code 2.1.305 added to /usage.

Terms in this piece · Glossary

AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
MCP — The Model Context Protocol — an open standard that lets any AI assistant plug into any tool or data source without custom integration code.
agent hook — A script the harness runs automatically at a fixed point in an agent's loop — before a tool call, after an edit, on session start.

Read the source github.com