Deep Agents: LangChain's SDK for Agents That Plan and Delegate - Talk Python to Me Ep. 543

When you type a question into ChatGPT, the model only has what you typed to work with. But tools like Claude Code can plan, iterate, test, and recover from mistakes. They work more like we do. The difference is the agent harness: Planning tools, file system access, sub-agents, and carefully crafted system prompts that turn a raw LLM into something genuinely capable. Sydney Runkle is back on Talk Python representing LangChain and their new open source library, Deep Agents: A framework for building your own deep agents with plain Python functions, middleware hooks, and MCP support. This is how the magic works under the hood. ▬▬▬▬ About the podcast ▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬ This video is the uncut, live recording of the Talk Python To Me podcast ( https://talkpython.fm ). We cover Python-focused topics every week and publish the edited and polished version in audio form. Subscribe in your podcast player of choice (100% free) at https://talkpython.fm/subscribe. ▬▬▬▬ Guests ▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬ Sydney Runkle ▬▬▬▬ Links and resources from the show ▬▬▬▬▬▬▬▬▬▬▬▬ Claude Code uses: https://x.com/alexalbert__/status/1948765443776544885?ref=blog.langchain.com Deep Research: https://openai.com/index/introducing-deep-research/?ref=blog.langchain.com Manus: https://manus.im/?ref=blog.langchain.com Blog post announcement: https://blog.langchain.com/deep-agents/ Claudes system prompt: https://github.com/kn1026/cc/blob/main/claudecode.md?ref=blog.langchain.com sub agents: https://docs.anthropic.com/en/docs/claude-code/sub-agents?ref=blog.langchain.com the quick start: https://docs.langchain.com/oss/python/deepagents/quickstart CLIs: https://github.com/langchain-ai/deepagents?tab=readme-ov-file#deep-agents-cli Talk Python's CLI: https://talkpython.fm/blog/posts/talk-python-now-has-a-cli/ custom tools: https://docs.langchain.com/oss/python/deepagents/overview#create-a-deep-agent DeepAgents Examples: https://github.com/langchain-ai/deepagents/tree/main/examples Custom Middleware: https://docs.langchain.com/langsmith/custom-middleware#how-to-add-custom-middleware Built in middleware: https://docs.langchain.com/oss/python/deepagents/customization#middleware Improving Deep Agents with harness engineering: https://blog.langchain.com/improving-deep-agents-with-harness-engineering/ Prebuilt middleware: https://docs.langchain.com/oss/python/langchain/middleware/built-in ▬▬▬▬ Dive deeper ▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬ Listen to the Talk Python To Me podcast at https://talkpython.fm and explore over 275 hours of Python courses at https://training.talkpython.fm/courses. ▬▬▬▬ Follow us on Social ▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬ Mastodon: Michael - https://fosstodon.org/@mkennedy & Talk Python - https://fosstodon.org/@talkpython Bluesky: Michael - https://bsky.app/profile/mkennedy.codes & Talk Python - https://bsky.app/profile/talkpython.fm X: Michael - https://x.com/mkennedy & Talk Python - https://x.com/talkpython