Give your agent access to
your authenticated browser
Stop paying for residential proxies and fighting CAPTCHAs. Conduit piggybacks on your active session, streaming authenticated data directly to your AI via MCP. A fully headless architectural integration.
Zero CAPTCHAs. Zero Proxy Costs.
Conduit gives your AI agents real-time browser context using your real, authenticated sessions — bypassing the pain of Playwright and traditional cloud scraping.
Passive Data Capture
Automatically intercepts XHR/Fetch API responses on supported platforms. Structures raw data into clean, typed JSON — without touching the DOM or breaking pages.
MCP + REST Server
19 MCP tools via Streamable HTTP (port 7779) or stdio for Claude Desktop and Cursor. Full REST API on port 7777 for scripts and dashboards. Passive context is completely free.
Autonomous Amazon Ingestion PRO
Import your entire Amazon order history automatically. Your AI agent can programmatically navigate, parse, and collect past orders with product images directly through the MCP tool.
Browser Automation PRO
Open tabs, navigate pages, click elements, type text, scroll, and capture screenshots — all remotely via API. Unlocks full active control over the browser.
Page Intelligence
Read any page's full text, metadata, Open Graph tags, and highlighted text via API. "Summarize the page I'm reading" is one free API call away.
Three steps to browser intelligence
No configuration files, no complex setup. Install and go.
Install the extension
Add Conduit from the Chrome Web Store. It runs silently in the background — no popups, no interruptions.
Browse normally
Conduit passively intercepts API responses on supported platforms. Shopping, social media, video — it structures everything automatically.
Connect your agent
Connect Claude Desktop, Cursor, or any MCP-compatible agent to stream real-time browser context directly into your reasoning loop. No GUI distractions.
20 Native MCP Tools
Conduit exposes a comprehensive suite of native browser controls directly to your AI.
🏆 Tier 1: The Sovereign Primitives
These tools represent the core "magic" of Conduit. They bypass traditional roadblocks (CAPTCHAs, headless detection) and give your agent unparalleled active leverage.
conduit_amazon_history
Import Amazon order history automatically via tab navigation and DOM extraction.
conduit_browser_action
The steering wheel. Execute a browser action on a tab: click, type, scroll, read DOM, or get URL.
conduit_read_page
The foundational tool for RAG. Read full text + images from a browser tab.
conduit_execute_js
The backdoor. Execute JavaScript in a browser tab to pull deep state (React/Redux caches).
conduit_browse_data
The data net. Read passively captured API response JSON from the browser (Amazon, Walmart, X).
🥈 Tier 2: The Core Navigators
These tools form the everyday scaffolding required for an agent to maintain spatial awareness and control over the browser session.
conduit_list_tabs
The agent's map. List all open browser tabs with IDs, URLs, and titles.
conduit_create_tab
The ignition spark. Open a new browser tab with the specified URL to start targeted research.
conduit_page_text
Token efficiency. Read strictly visible text + images from a browser tab without raw HTML bloat.
conduit_screenshot
Visual grounding. Capture a Base64 screenshot of the currently visible browser tab for VLM models.
conduit_get_cookies
Instant identity. Get cookies for a domain, including HttpOnly tokens, for cloud worker handoffs.
🥉 Tier 3: Tactical Utilities
Hyper-specific tools used in edge cases or highly structured data engineering loops.
conduit_get_selection
Get highlighted/selected text actively chosen by the human user in the browser.
conduit_page_meta
Fast orientation. Get tab metadata: title, URL, OG tags, Twitter card, favicon.
conduit_drain_data
Get captured events AND remove them from the buffer (consume-once queue implementation).
conduit_keypress
Fidelity fallback. Simulate strict key presses (Enter, Tab, Esc) for stubborn UI controls.
conduit_history
Forensic search. Query persistent SQLite footprint history by platform or time range.
🛠️ Tier 4: Lifecycle & Garbage Collection
conduit_close_tab
Close an existing browser tab to sweep up after autonomous workflows.
conduit_page_images
Collect image URLs specifically from recent captured network events.
conduit_clear_data
Manually flush the event buffer.
conduit_history_stats
Get SQLite persistent store sizes and limits.
conduit_status
Ping extension bridge status: connection state, buffer size, uptime.
One Extension. Two Tiers.
Conduit silently captures your structured data via a headless architecture. Upgrade to dev_pro to unlock persistent history and MCP integration.
- Headless background capture
- Local MCP Server Integration
- Read-Only REST API
- Page Intelligence (DOM Extraction)
- Amazon & Walmart parsing
- Everything in dev_free
<all_urls>access- Browser automation API (Click, Type)
- Autonomous Amazon Ingestion
- WebSocket content streams
- Cookie extraction API
Smart data extraction, built in
Each connector understands the platform's data structure and extracts clean, typed JSON automatically. No configuration needed.
Amazon
Orders, products, pricing, reviews
ACTIVEWalmart
Products, pricing, availability
ACTIVE
All connectors included in every edition — Consumer and Pro.
+ more platforms shipping regularly
What's coming next
Conduit is actively developed. Here's what we're building.
Proactive Agent Subscriptions
Shift from polling to event-driven architectures. Subscribe your agent to DOM mutations or network events (e.g., price drops) to wake up the reasoning loop instantly.
IN DEVELOPMENTSemantic Browsing Memory
Conduit automatically embeds intercepted JSON payloads into a local vector database. Your agent can perform semantic search across the user's entire historical browsing context.
IN DEVELOPMENTOmniCommerce Write Orchestration
Move beyond read-only intercepts. A unified E-commerce API to allow your agent to autonomously add items to universal carts across Amazon and Walmart with human-in-the-loop approval.
COMING SOONDeep State Extraction
Bypass brittle DOM selectors. Conduit will extract structured app state directly from React Fiber trees and Apollo caches, feeding raw state fragments to your agent.
COMING SOONYour data never leaves your computer
Conduit processes and stores everything locally. No external servers, no cloud uploads, no tracking. The captured data lives on your machine and is served only to localhost. You own your data completely.
