Full-stack observability startup Hound Technology Inc., which does business as Honeycomb, introduced a number of new platform updates aimed at investigating artificial intelligence agent activity in production. 

The new enhanced capabilities provide deeper visibility into what AI agents are doing while they’re running, the company said. The enhanced capabilities include Agent Timeline, Canvas Agent and Canvas Skills. Engineering teams can enable them without the need for proprietary software development kits or specialized frameworks. 

“AI agents are now part of the engineering team,” said co-founder and Chief Executive Christine Yen. “But right now, most teams can’t see what those agents are doing in production: which tools they called, what they decided, whether they made things better or worse.” 

Honeycomb said its new enhanced capabilities aim to change this dynamic between agents and information technology teams. Yen added that Honeycomb is taking its already existing expertise in human-centric observability design and using it to track the behavior of agents, which are provided roles and tool use paradigms similar to human counterparts. 

The company said the new updates surface analysis for AI systems including behavior, performance and the interactions of AI agents, alongside deeper visibility into systems using AI itself. 

Agent Timeline provides a single view that connects every large language model call, agent handoff and tool invocation – such as email view, opening a text editor, calling the content management system and so on. The dashboard allows users to visualize downstream system impact in real time, allowing engineers teams to trace activity, reconstruct agent decision paths and understand failures without needing to make manual deep dives into logs. 

Honeycomb newly rebuilt Canvas, the company’s AI-human collaborative workspace, which now acts as a chat interface and autonomous agent all in one. It allows engineering teams to plunge investigate observability issues using plain English queries and produces visual snapshots of system activity. 

Boosting the capabilities of Canvas, the company added Canvas Skills, which allow the engineering teams to teach the AI agents routine and best practice debugging knowledge as reusable playbooks that can run autonomously. This means that when a similar issue arises in the future, engineers don’t need to write out long explanatory prompts. The AI agent uses this taught knowledge as a foundation to explore and investigate, freeing up the engineering team to ask more pointed questions about discoveries. 

“Honeycomb was built to tackle the ‘unknown unknowns,’ those unforeseen failures that defy planning.” said Yen. 

Honeycomb also added auto-investigations, which allows engineers to set Canvas to automatically begin investigations when an alert arrives. Instead of waiting for the team to ask questions, it will run playbooks against anomalies, gather data, create and test hypotheses and suggest responses, all before the engineering team even reaches their terminals. 

The company said Agent Timeline is out now in early access and will be generally available in a few weeks. The other updates are immediately available to customers starting today. 

 Image: SiliconANGLE/Microsoft Designer