Claude Agent SDK: Long-Running Agents with Cole Medin | AMA

Description

Cole Medin walks us through Anthropic's latest engineering blog on building effective harnesses for long-running agents. We're using Opus 4.5 with the Claude Agent SDK to actually build this out live. The core problem: Agents lose all context between sessions. Anthropic's solution uses two specialized agents - an initializer that sets up the environment, and a coding agent that makes incremental progress while leaving clean artifacts for the next session. What we're covering: - The initializer agent: feature lists, init.sh scripts, progress tracking files - The coding agent: incremental work, git commits, structured updates - Why JSON feature files work better than Markdown for agent instructions - Browser automation testing with Puppeteer MCP - Real implementation using the Claude Agent SDK quickstart AMA throughout - bring your questions about multi-context window workflows. Resources: Anthropic Engineering Article: https://www.anthropic.com/engineering/effective-harnesses-for-long-running-agents Claude Agent SDK Quickstart: https://github.com/anthropics/claude-quickstarts/tree/main/autonomous-coding Claude Agent SDK Docs: https://platform.claude.com/docs/en/agent-sdk/overview [TIMESTAMPS]