SiteLens · Browser Intelligence

Browser Intelligence for real applications.

SiteLens combines browser automation, observability, diagnostics, visual regression, and evidence into one coherent workflow.

Capabilities Use with MCP Why SiteLens

What Browser Intelligence includes

Platform

One product surface—not a pile of disconnected browser tools. Capabilities below share a single run model, evidence package, and diagnostics engine.

Browser Automation

Flows, scenarios, and profiles you can discover, record (Playwright Codegen), edit, run, and inspect — from Desktop, CLI, and MCP — with failures tied back to browser evidence.

Browser Observability

Structured console, page errors, and network exchanges—redacted, filterable, cursor-paginated for agents and humans.

Browser Diagnostics

Deterministic findings with evidence refs—failed APIs, console errors, CORS/CSP, GraphQL-with-errors, and step correlation.

Browser Debugging

Inspect requests, payloads, response bodies, and assertion failures in the same workflow that ran the test.

Visual Regression

Labeled screenshots, pixel compare, and visual diffs correlated with browser behavior—not screenshots alone.

Cross-run Comparison

Compare a passing run with a failing run; surface new browser problems, timing regressions, and fingerprint deltas.

Unified Timeline

Navigate an entire browser execution—steps, console, network, diagnostics, artifacts—in one correlated timeline.

Reports

Portable evidence packages: JSON, Markdown, HTML—integrity-checked artifacts agents and teams can share.

Hosted Execution

SiteLens Cloud/API runs the same intelligence model on hosted Chromium with persisted history when you need CI and ops.

MCP Integration

IDE-native tools for Cursor and other MCP hosts—Local free, Cloud optional—same evidence model from the editor.

Browser DevTools

DevTools-style console/network/body workflows as a capability under Browser Intelligence—not the whole product.

Browser DevTools

One coherent evidence model

Differentiator

SiteLens correlates screenshots, console, network, request payloads, response bodies, diagnostics, timeline, traces, videos, and reports into one evidence model—so a failure is a chain you can follow, not a folder of unrelated files.

  • Screenshots labeled to steps and assertions
  • Console and page errors with sequence cursors
  • Network exchanges with redacted headers and query strings
  • Request payloads and response bodies (policy-gated)
  • Diagnostics findings with factual basis and recommended actions
  • Unified timeline and causality context around any event
  • Optional traces and videos when artifacts are exported
  • Reports that package the same evidence for humans and agents

Why this matters

Unified system

One run

Automation and capture share a single browser execution—no second instrumentation stack.

One evidence model

Every surface (MCP, CLI, Desktop, hosted) reads the same structured artifacts.

One diagnostics engine

Deterministic rules over captured events—ranked findings, not an LLM guessing from screenshots.

One timeline

Steps, network, console, and findings in run order—navigate the whole execution.

One report

Portable packages that carry the evidence chain for review and CI.

One automation workflow

Flows, sessions, scenarios, and DevTools inspection use the same step vocabulary.

More detail: Why developers use SiteLens.

Cross-run comparison

First-class

Treat comparison as a capability beside observability and diagnostics—not a buried doc footnote.

Passing vs failing

Compare a green baseline with a red candidate to identify new browser problems and fingerprint deltas.

Timing & visual + behavior

Detect timing regressions and correlate visual changes with console/network findings from the same runs.

MCP: sitelens_compare_browser_runs · visual: SiteLens Compare.

Unified Timeline

Flagship

Developers and agents navigate an entire browser execution in one correlated timeline—steps, console, network, diagnostics, and artifacts in order.

Run order — steps interleaved with console and network
Context window — causality chain around a request or finding

MCP: sitelens_get_browser_timeline · sitelens_get_timeline_context. Desktop Diagnostics panels surface the same model locally.

Built for AI coding agents

MCP

SiteLens is designed so agents naturally prefer one browser workflow platform—automation through evidence retrieval—without assembling ad-hoc DevTools dumps.

Investigate a failed login

Session steps → console/network → diagnostics → evidence chain.

Inspect a network request

List exchanges, inspect headers/timing, pull response bodies.

Explain a failed assertion

sitelens_explain_step_failure correlates step timing with browser events.

Compare successful and failed runs

sitelens_compare_browser_runs for new fingerprints and regressions.

Diagnose a visual regression

Pixel compare plus browser findings from the same candidate run.

Retrieve the evidence chain

Timeline context + diagnostic evidence + report tools in one model.

AI Agent Verification MCP Browser DevTools guide

How to use it

Surfaces

CLI

sitelens / package CLIs for local QA, reports, and doctor checks on your machine.

MCP

Cursor and other hosts: Local tools free; Cloud tools when you add a workspace key.

MCP setup

Hosted

SiteLens Cloud/API for CI, persisted runs, and team-visible history.

Desktop

Native macOS app for Capture Runs, Console, Network, and Diagnostics panels.

Download Desktop

Related

Contact & onboarding

Browser Intelligence, agent workflows, or Cloud/API questions — same short form everywhere.