BrowserLane Inspector
A local-first desktop app for reading the forensic traces bl records — see what an agent did in the browser and what broke, without anything leaving your machine.
Coming soon
BrowserLane Inspector is in active development and is not yet available
for download. This page describes what's coming. The capture side — the
forensic trace recorder — ships in bl today.
What it is
BrowserLane Inspector is a desktop app for reading the forensic traces bl
records. When an agent drives the browser, the trace captures what actually
happened: every command it issued, and — at the debug and forensic capture
levels — the screenshots, console lines, and network requests of the session
run. Inspector turns that folder of events into
something you can read — a timeline of what the agent did, the evidence behind
each step, and an answer to "what broke?" — with nothing ever leaving your
machine.
It is a strictly read-only viewer: it doesn't drive the browser, start or
stop sessions, or modify trace contents — the only write operation is
explicitly deleting a trace. bl writes; Inspector reads.
Key features
- Timeline of tool-call steps — events are grouped into tool-call steps, with console output, network requests, and screenshots attributed to the call they occurred under (best-effort — evidence outside a call's window appears unattributed). Every step shows its duration and status, and failed steps auto-expand.
- Failure triage ("What broke?") — the last failing tool call, the sequence from last-ok to failed, related console errors and failed/5xx requests, and the last screenshot before the failure — each one a click-through to its event.
- Live tail — a running session's timeline updates as events append, so you can watch a trace grow while the agent works.
- Network & console tabs — a DevTools-style request table (method, status, type, host, path, duration) with request-to-response correlation, failed-only and resource-type filters; a console view with level filters and page errors flagged.
- Screenshot filmstrip & artifacts — step through screenshots one at a time, jump straight to the first failure, or browse everything the run captured in a grid.
- Run browser — search, status and time filters, and sorting across all recorded runs; repeated runs of one session name group as "run k/n", and running sessions get a pulsing indicator.
- Event inspector — sanitized arguments, a result preview, redacted headers, and the raw JSON for any event.
- Trace management — reveal a trace in Finder, export it as a
.zipbundle, or delete it from the app.
How it works with bl
Traces come from the recorder built into bl: set
BL_TRACE=normal|debug|forensic and each session run is recorded as one
folder — session metadata, an append-only event log, and an artifacts
directory (see the tracing guide). Inspector watches the
traces directory: new runs appear as they start, and a running run's timeline
fills in live as bl appends events.
Privacy
- Local-only by design. The app makes zero network requests and has no telemetry.
- Redaction happens at capture time, in
bl, before anything hits disk: sensitive key names, JWT and card-number patterns, and the values typed into sensitive fill targets are masked. Inspector surfaces each trace's capture level and redaction status in the UI. - Built with Tauri v2 — a Rust backend and a React front end, nothing cloud-hosted.
Availability
Inspector is in active development; availability will be announced on
browserlane.com. Until then, the trace format is
documented (schema v1) and traces are plain JSONL — you can already read
today's recordings with a text editor and jq.