Europe 2026
Building Agent Interfaces: Lessons from Chrome DevTools (MCP) for Agents — Michael Hablich, Google
Overview
This talk introduces Chrome DevTools for Agents, a purpose-built debugging tool for AI agents, drawing parallels to the widely used Chrome DevTools for human web developers. The core thesis is that agents, as a distinct user class, require specialized interfaces and debugging capabilities to overcome their limitations, such as processing vast amounts of data or understanding complex tool descriptions. The presentation emphasizes engineering lessons learned from building and deploying this tool, focusing on effectiveness, efficiency, discoverability, and trust.
Who should watch
- AI engineers building or integrating agentic workflows.
- Product Managers seeking to understand agent capabilities and limitations.
- Developers working with tool-calling or function-calling models.
- Anyone interested in improving agent reliability and performance.
- Those facing challenges with agent data processing and decision-making.
Key takeaways
- Agents, like humans, have different cognitive bottlenecks; humans are visual, while agents struggle with large data volumes and nuanced tool descriptions.
- Measure agent interface efficiency using metrics like tokens per successful outcome, but always balance this with effectiveness to ensure the agent reaches its destination.
- Improve agent error recovery by providing clear error messages and implementing diagnostic playbooks to enable self-healing.
- Enhance tool discoverability by clearly defining tool purpose and providing usage guidelines, as poor descriptions can lead agents to misuse tools.
- Address token burn by categorizing tools, offering slim modes, and enabling command chaining for post-processing tasks.
- Prioritize trust boundaries in agent interfaces, especially for agents with internet access, by implementing measures like consent for actions and domain allow-listing.
- The schema, or tool descriptions, acts as the UI for agents; improving these descriptions is crucial for effective agent interaction.
- Agent experience is an evolution of user experience, requiring consideration of non-functional requirements like efficiency, discoverability, and security.
Notable quotes
*We ship Chrome DevTools because we saw that coding agents were flying blind.*
*Essentially, we didn't force the agent to read the entire book, the trace, but instead we just pointed it at the right sentence.*
*Fuel efficiency is relatively worthless if you can't reach your destination.*
Unofficial community note. Prefer the recording for nuance.