claude-agent-sdk-typescript v0.3.304
- Source
- github.com
- Author
- ashwin-antTop Viber
- Date

Context
The most practical change in version 0.3.304 of Anthropic's TypeScript SDK is a fix to `startup()`, which could time out simply because a startup was still running (release notes). `startup()` pre-warms an agent: it launches the Claude Code process the SDK drives and completes the opening handshake before any prompt exists, so the first real query does not pay that startup cost. It waits up to `initializeTimeoutMs` for that, 60 seconds by default, and rejects with a timeout error if initialization takes longer (SDK reference).
The problem was hooks. SessionStart and Setup command hooks are commands configured to run as a session begins, and while they ran, their time counted against that 60-second limit, so a slow hook could make `startup()` time out. According to the notes, when no `initializeTimeoutMs` is passed, up to 10 minutes of hook time no longer counts toward the limit, and a timeout error now includes what the process reported, which makes the cause easier to find. The notes describe the change only for that default case; they do not say how an explicitly set timeout treats hook time. A related fix stops `startup()` from holding every progress message from a running hook in memory and keeps only the newest one per hook.
- 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.
- 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.
- agent skill — A reusable instruction file that teaches an agent how to do one job well — the procedure, the tools, and what counts as done.
- MCP — The Model Context Protocol — an open standard that lets any AI assistant plug into any tool or data source without custom integration code.
Checking sign-in…
Loading comments…
