Foxglove adds AI agents, semantic search, and A/B comparison for robotics data

Foxglove, the data platform for robotics developers, announced a major update centered around AI agents and smarter data workflows. The search experience now merges traditional topic-based queries with semantic search, enabling operators to find specific moments using natural visual cues and numeric thresholds. For example, `visual("stairs") AND (/anymal/imu/linear_acceleration.x > 4 OR /anymal/imu/linear_acceleration.x < -4)` returns instances of a robot walking on stairs that match both a visual concept and a sensor signature. Clicking any result jumps directly to that exact moment in the recording.

The centerpiece is the new Agent Sidebar. Users can ask an AI assistant to create layouts, triage issues, write userscripts, or generate events without leaving the app. Foxglove also added an MCP (Model Context Protocol) interface in the desktop app, allowing any standards-compliant agent to be connected and interact with the workspace. One suggested prompt: ask the agent to triage an issue and produce a PDF report. Beyond agents, Foxglove introduced A/B comparison, which aligns time offsets across two or more runs so teams can visualize odometry performance before and after calibration on the same timeline. Also new: Remote Access streams from a robot via WebRTC with adaptive quality, and Bring Your Own Server supports external storage buckets to eliminate duplicate data copies.