v2.0 — 12 MCP tools • Persistent sessions • Page comprehension

Your AI agent's browser

Navigate, authenticate, interact, observe, and extract from any web page. Anno gives AI agents a body on the internet — with stealth browsing, persistent sessions, and 93% token reduction.

Get Your API Key Try It Live

AI Agents Can Think. They Can't Browse.

Every AI browser tool gives agents hands. None of them give agents understanding.

🔒

Can't authenticate

Cloudflare blocks them. Login forms confuse them. Sessions die between calls. Your agent is locked out of 90% of the web.

🤷

Can't comprehend

They get raw HTML or markdown. They don't know if it's a login page, a product page, or a captcha. They click the wrong button because they never understood the page.

💸

Can't be efficient

15,000 tokens of HTML nav, ads, and scripts to get 600 tokens of content. You're paying for noise. 93% of your token budget is wasted.

One API. Complete Web Autonomy.

Multi-step authenticated browsing in 4 tool calls

Step 1

🔑

Authenticate

anno_session_auth
Solve Cloudflare. Get cookies. Create a persistent session.

Step 2

👁

Observe

anno_observe
Classify the page. Find interactive elements. Detect patterns.

Step 3

🖱️

Interact

anno_interact
Click, fill, scroll, type. Same session. Same cookies.

Step 4

📄

Extract

anno_fetch
Confidence-scored content. 93% token reduction. Provenance tracked.

See the Difference

97% median token reduction — benchmarked across 22 production sites

Without Anno (Raw HTML) ~15,200 tokens

<!DOCTYPE html> <html lang="en" data-edition="us" class="pg-type-homepage"> <head> <meta charset="utf-8"> <meta name="viewport" content="width=device-width"> <title>CNN - Breaking News, Latest News...</title> <link rel="stylesheet" href="/css/bundle-47af2e.css"> <script src="/js/analytics-tracking.min.js"></script> <script src="/js/ads-prebid-config.js"></script> <script>window.__CNN_CONFIG__={...}</script> ... 14,800 more tokens of scripts, styles, nav, tracking pixels, ad containers, cookie banners, A/B test configs, social widgets, modals, lazy-load placeholders, inline SVGs ... <div class="cd__headline-text"> Actual headline buried somewhere in here </div> </html>

Tokens: ~15,200 Content: ~5% of payload

With Anno ~620 tokens

{"type":"title","content":"CNN Breaking News"} {"type":"heading","content":"Senate passes bill..."} {"type":"text","content":"The Senate voted 52-48 on Thursday to pass the controversial spending bill, sending it to the House for a final vote expected next week."} {"type":"heading","content":"Markets rally on Fed signals"} {"type":"text","content":"Wall Street surged after Federal Reserve Chair indicated rate cuts could begin as early as September, with the S&P 500 closing up 1.2%."} {"type":"heading","content":"Weather: Major storm system"} {"type":"text","content":"A powerful storm system is expected to bring heavy rain and flooding to the Southeast this weekend, with up to 6 inches of rain forecast."}

Tokens: ~620 Content: 100% of payload

Try It Live

Paste any URL and see Anno extract the content in real-time

Click "Extract" to see Anno in action...

Three Lines to Clean Web Content

Works with any language. REST API or MCP tool — your choice.

# Extract clean content from any URL curl -X POST https://anno.evolvingintelligence.ai/v1/content/fetch \ -H "Authorization: Bearer YOUR_API_KEY" \ -H "Content-Type: application/json" \ -d '{"url": "https://example.com", "render": false}'

12 MCP Tools. One Surface.

Works with Claude Code, Cursor, or any MCP-compatible agent

🔑

Navigate & Authenticate

anno_session_auth New

Solve Cloudflare challenges, inject cookies, return authenticated session with persistent sessionId
anno_interact New

Click, fill, scroll, hover, type — act on any page through a stealth browser
anno_workflow New

Multi-step browser automation with conditionals, loops, and variable interpolation

👁

Observe & Understand

anno_observe New

Page comprehension — classify page type, find interactive elements, detect captchas/paywalls/popups
anno_page_state New

Inventory every button, link, input, and select on the page with CSS selectors
anno_screenshot New

Visual capture returned as MCP image content — see what your agent sees

⚡

Extract & Search

anno_fetch

4-method extraction ensemble with confidence scoring — 93% token reduction vs raw HTML
anno_batch_fetch

Up to 10 URLs in parallel with per-URL confidence scoring
anno_crawl

Site-wide discovery with BFS/DFS, robots.txt compliance, and depth limits
anno_search New

Semantic search over previously extracted content via vector similarity

📡

Monitor & Manage

anno_watch New

Monitor URLs for content changes — semantic diffing with configurable thresholds and webhooks
anno_health

Server status, cache stats, browser availability

What Nobody Else Does

Browser Use gives agents hands. Firecrawl gives agents markdown. Anno gives agents understanding.

Page Comprehension

anno_observe classifies page type (login, product, article, checkout), detects patterns (captcha, paywall, popup), and maps interactive elements — before your agent acts.

Persistent Sessions

Authenticate once, browse everywhere. sessionId threads across all tool calls — your agent maintains cookies, state, and context across multi-step workflows.

Confidence Scoring

Every extraction is scored. Your agent knows how reliable the content is — not just what was extracted, but how much to trust it.

Provenance Tracking

SHA-256 content hashes, byte-level source spans. Know exactly where every piece of content came from. Auditable by default.

Cloudflare Bypass

Stealth Playwright browser with anti-detection, canvas fingerprinting protection, and realistic profiles. Solves challenges that block every other tool.

93% Token Reduction

14 web pages in the space raw HTML uses for 1. Not markdown approximation — structured, confidence-scored nodes optimized for LLM consumption.

Built For

Real problems developers are solving right now

AI Agents That Need Authenticated Data

Price monitoring behind login walls. Account data export. Competitive intelligence on protected pages. Anno authenticates, maintains the session, and extracts — your agent just asks.

RAG Pipelines Eating Token Budgets

You're feeding 15K tokens of HTML per page into your retrieval pipeline. Anno returns 600 tokens of structured content. Same information, 1/25th the cost. Batch 10 URLs in parallel.

Web Automation That Needs to Be Reliable

Browser Use tries to figure out each page autonomously and fails on page 2. Anno's workflow engine executes multi-step sequences deterministically — with conditionals, loops, and variables.

Marketplace Intelligence

Typed extraction adapters for eBay sold listings, Amazon products, and Walmart pricing. Field-level data with confidence scores. Built for FlipIQ-style resale research.

Built for Production

API Key Auth

SHA-256 hashed keys, multi-tenant isolation, Bearer token and X-API-Key support

Rate Limiting

Token bucket algorithm with per-key limits and standard headers

Redis Caching

Configurable TTL, cache-aware pipelines, automatic deduplication

Prometheus Metrics

Request counts, latencies, cache hit rates, extraction quality scores

MCP Native

12 tools for Claude Code, Cursor, or any MCP-compatible agent framework

NDJSON Streaming

Real-time event stream — metadata, confidence, nodes, provenance in one call

Simple Pricing

Start free, scale as your agents grow

Free

$0 /month

For individual developers and experimentation

✓ 200 requests/month
✓ All 12 MCP tools
✓ Stealth browser rendering
✓ Persistent sessions
✓ Community support

Get Free API Key