The AWS Developers Podcast Mike Chambers: From OpenClaw to AI Functions — What's Next for Agentic Development
11 snips
Feb 25, 2026 Mike Chambers, Senior Developer Advocate at AWS who builds agentic systems and open-source tooling. He unpacks OpenClaw’s rise and why local routing frameworks matter. He walks through async tool calling that keeps conversations flowing while long tasks run. He explores Strands Agents SDK, AI Functions as a new runtime idea, and what observability and trust mean for future agentic software.
AI Snips
Chapters
Transcript
Episode notes
Running OpenClaw On A Mac Mini With Mixed Models
- Roman runs OpenClaw on a Mac mini and mixes local LLMs with cloud models depending on task complexity.
- He keeps sensitive personal interactions local and experiments with voice (Whisper, Novasonic) and Llama/Llama.cpp for privacy and latency trade-offs.
Asynchronous Tool Calls Keep Conversations Natural
- Asynchronous long-running tool calls let agents continue a natural conversation while background tasks run.
- OpenClaw and Pi-style frameworks spawn sub-agents and can aggregate or delay responses to avoid interrupting user flow.
Implement Async Tools With Immediate Acknowledgement
- Do use an asynchronous tool decorator to mark long-running tools and return immediately with a unique reference ID.
- Store the job in memory and re-insert results into the conversation only when the current turn completes to avoid cutting users off.