1-line AI SDK integration
Drop recordStep into the Vercel AI SDK’s onStepFinish and every step is
captured. agentvu never imports the ai package — it matches the step’s
shape, so it stays dependency-light.
agentvu
Watch your AI agent think, live, in a beautiful terminal dashboard. agentvu is a tiny library + CLI that records an AI agent’s run as a stream of structured events and visualizes it as a timeline of model calls, tool calls (with their args and results), tokens, cost, and durations. Replay a finished run, or follow it live as it happens.
An agent run is invisible by default. The model fires off a tool call, gets a result, reasons again, calls another tool, and finally answers — but all you see in your logs is a wall of JSON, if anything. agentvu turns that run into a stream you can actually read: one structured event per line, rendered as a dashboard you can replay or watch in real time.
agentvu session.jsonlsteps 3 tools getWeather,getForecast tokens 1996 cost $0.02 dur 3.0s
▍ user What's the weather in Tokyo and should I bring an umbrella?◆ gpt-4o system + user: weather question for Tokyo (820ms)→ getWeather {"city":"Tokyo","units":"metric"}← getWeather {"tempC":18,"condition":"light rain",...} (260ms)∑ usage 412 in / 86 out (gpt-4o)◆ gpt-4o reason over weather tool result (640ms)→ getForecast {"city":"Tokyo","hours":6}← getForecast {"chanceOfRain":0.72,"summary":"rain likely..."} (230ms)∑ usage 540 in / 120 out (gpt-4o)◆ gpt-4o compose final answer (710ms)▍ assistant It's 18°C with light rain in Tokyo and a 72% chance...∑ usage 680 in / 58 out (gpt-4o)Zero lock-in
Events are plain JSON, one per line (JSONL). No service, no proprietary SDK
to learn, no network call. Tail it, grep it, commit it — it’s just a file.
Live or replayed
Render a finished run, or pass --follow to tail the file and re-render the
dashboard every time your agent appends a new event.
Cost & tokens built in
An approximate, built-in price table for popular models turns token usage into a ballpark dollar figure right in the header.
Next steps
Section titled “Next steps” Quick start From install to a recorded and visualized run in a few steps.
Recording The event model, the recorder, the AI SDK adapter, and cost estimation.
CLI The agentvu command — arguments, flags, exit codes, and examples.
API The exported library surface — functions and key types.