Documentation

All Docs

Browser Inspector

51 MCP tools that turn your browser into an AI-controllable DevOps platform.

What is the Browser Inspector?

The Browser Inspector is a bridge between the Neuron extension and AI coding agents like Claude Code, Cursor, or Windsurf. It exposes 51 MCP tools that let an AI agent inspect, automate, debug, audit, and extract data from any page in your browser — all through natural language. Think of it as giving your AI assistant full DevTools access.

Set it up

Three steps to connect your browser to Claude Code:

  1. 1

    1. Install the extension

    Download and install the Neuron extension from neuron.ng/extension. Load it in Chrome via Developer Mode → Load unpacked.

  2. 2

    2. Build the bridge

    In your Neuron project directory, run: make inspector-setup — this builds the bridge server and registers it with Claude Code as an MCP server.

  3. 3

    3. Start the bridge

    Run: make inspector — this starts the MCP server (stdio) and a WebSocket server on localhost:7377. The extension connects automatically when dev mode is enabled.

You can also register manually: claude mcp add neuron-inspector -- node packages/extension/bridge/dist/server.js

What it can do

The 51 tools are organized into 10 categories. Here is what each category covers:

  1. 1

    Inspect & Debug

    DOM snapshots, element search, JavaScript evaluation, network request history, console logs (all levels), WebSocket frames, HAR export, and screenshots.

  2. 2

    Browse & Automate

    Click elements, type into inputs, navigate and scroll pages, open and reload tabs, run multi-step action sequences, and trigger Instagram posts.

  3. 3

    Security Scan

    Detect leaked API keys (AWS, Stripe, Firebase, GitHub, JWT), audit security headers (CSP, HSTS), check CORS config, find mixed content, insecure forms, and exposed source maps.

  4. 4

    Performance & Quality

    Core Web Vitals (LCP, CLS, FCP, TTFB), network waterfall with timing breakdown, resource analysis, WCAG accessibility audit, and SEO audit (meta tags, OG, headings, structured data, word count).

  5. 5

    Traffic Intelligence

    Search all response bodies by string or regex, auto-discover and document APIs from captured traffic, replay requests with modified parameters, and extract structured data from any page.

  6. 6

    Network Mocking

    Mock any URL pattern with custom status codes and response bodies, simulate slow APIs with delays, test error handling without touching the backend.

  7. 7

    Cookies & Storage

    Read, set, and delete cookies. Inspect and clear localStorage and sessionStorage. Check cookie expiry and security flags.

  8. 8

    Page Monitoring

    Watch any CSS selector for text, visibility, or attribute changes over time. Snapshot full page state and diff two snapshots to detect what changed.

  9. 9

    Demo Recording

    Record user interactions as replayable workflows. Replay with cinematic overlays — chapter cards, captions, and a synthetic cursor. Export to MP4 or WebM.

  10. 10

    Self-Healing Agents

    Rate limit and captcha detection across Instagram, X, TikTok, LinkedIn, and Facebook. Automatic exponential backoff. Auth expiry diagnosis. Post-share API verification that catches silent failures.

Tools by plan

Basic inspection tools are free. Power tools unlock with your Neuron subscription:

  1. 1

    Free — 9 tools

    DOM queries, element search, screenshots, tab management, diagnose, navigate, open/reload tabs, scroll.

  2. 2

    Daily — 21 tools

    Everything in Free, plus: JS evaluation, network/error/console/WebSocket logs, click, type, action sequences, auth checks, blocker detection, session diagnostics.

  3. 3

    Pro — 39 tools

    Everything in Daily, plus: HAR export, network waterfall, performance snapshot, network mocking, cookies and storage management, workflow recording and replay, page state snapshots and diffing, element watching, post trigger.

  4. 4

    Team — 45 tools

    Everything in Pro, plus: response body search, API discovery, structured data extraction, request replay, SEO audit, accessibility audit.

  5. 5

    Business — 51 tools

    Everything in Team, plus: security scanning, demo video recording with cinematic overlays.

Local development builds unlock all 51 tools regardless of your plan — gating only applies in production.

Privacy and security

The inspector works on any website you browse — there are no domain restrictions. The bridge server itself binds to 127.0.0.1:7377, which means only programs on your own machine can talk to it (no one on your network can reach it remotely). There are no cloud services and no telemetry. Sensitive headers (authorization, cookies, CSRF tokens) are automatically redacted from all tool responses. Capture is opt-in per tab. Response bodies in HAR exports are excluded by default. The security scan detects secrets in your pages but never sends them anywhere.

Inspector vs Neuron MCP Server

Neuron has two MCP integrations that work together but serve different purposes. The Neuron MCP Server (mcp.neuron.ng) is a cloud service — it connects AI agents to the Neuron platform API for managing bots, conversations, broadcasts, contacts, and knowledge bases. No browser needed, works from anywhere. The Browser Inspector is a local bridge — it connects AI agents to your Chrome browser for inspecting pages, automating clicks, running audits, and extracting data. It requires the extension running on your machine. You can use both in the same Claude Code session. Register them separately: the platform MCP for backend operations, the inspector for browser operations. See the MCP Server docs for the platform integration.

Common use cases

A few things people use the inspector for:

  1. 1

    Debug a web app

    Ask your AI agent to inspect the DOM, read console errors, check network requests, and take screenshots — all without leaving your coding session.

  2. 2

    Audit before shipping

    Run security, accessibility, SEO, and performance audits in one conversation. Get structured findings with severity levels.

  3. 3

    Test error handling

    Mock API endpoints to return 500s, timeouts, or custom error bodies. Verify your UI handles every failure path.

  4. 4

    Reverse-engineer an API

    Browse a web app, then ask the agent to discover all APIs it called, show response shapes, and replay specific requests with different parameters.

  5. 5

    Extract data from a page

    Pull product listings, search results, or social feeds into structured JSON. Auto-detects repeating patterns or accepts a CSS selector.

  6. 6

    Record a product demo

    Record a workflow, replay it with chapter cards and captions, and export to MP4 — all driven by the AI agent.