LangSmith launches Custom Apps to tailor agent data views

LangChain has launched Custom Apps within its LangSmith developer platform, enabling engineering teams to design, publish, and run bespoke user interfaces on top of their LLM agent data. Instead of relying solely on default dashboards for metrics like latency and error rates, developers can now build tailored UIs to review traces, run experiments, and collect feedback. The feature handles hosting, authentication, and permissions automatically, removing the maintenance overhead typically associated with building custom internal tools.

Users can generate these interfaces using natural language prompts through LangSmith Chat, which translates descriptions of desired data and visualizations into functional applications. Alternatively, developers can write code using provided templates and the LangSmith API. By default, Custom Apps utilize Macaw, LangSmith's newly open-sourced design system, to maintain visual consistency, though teams can customize the styling to match their own branding.

For practitioners, this capability simplifies workflows like human-in-the-loop evaluation and experiment comparison. For instance, teams can build simplified annotation queues for non-technical subject matter experts while preserving detailed trace views for engineers. Roxanne Wang of Chime noted that the tool helped transition their evaluations away from scattered spreadsheets, creating a "repeatable framework for human-in-the-loop review" directly connected to the LangSmith API.

The new feature is available to customers on LangSmith's paid tiers. The Plus plan includes one Custom App per organization along with the chat creation experience. Enterprise plan subscribers receive unlimited Custom Apps alongside the chat interface.