Notte
🌸 Best framework to build web agents, and deploy serverless web automation functions on reliable browser infra.
Last verified:
What is Notte?
Notte is a unified browser infrastructure platform for AI agents and automation scripts that need to use the browser. It provides cloud browser sessions, browser agents, serverless browser functions, credential vaults, digital personas, and scraping capabilities all through a single API. Build, debug, and deploy production workflows with cloud browsers, web agents, scraping, serverless functions, credentials, and identities in one place.
Key features include stealth browser sessions with built-in CAPTCHA solving, proxies, and anti-detection; hybrid workflows that combine scripting and AI agents to reduce costs by 50%+ while improving reliability; enterprise-grade secrets vaults for storing emails, passwords, MFA tokens, and SSO credentials; and digital personas with unique emails, phones, and automated 2FA for account creation workflows. The platform offers 1000+ concurrent browser instances, <50ms latency, 99.9% uptime, >90% agent success rate, and works with any LLM model.
Notte is built for developers, product teams, and enterprises who need to automate web tasks at scale. It's perfect for web scraping, automated testing, form filling, data extraction, competitor monitoring, account creation workflows, and any task requiring browser automation. The platform is trusted by 100+ AI companies and partners, and is YC S25 backed. It supports Playwright, Puppeteer, Selenium, Python, TypeScript, Node.js, Docker, and integrates with OpenAI, Anthropic, Gemini, LangChain, Claude Code, and Cursor.
Notte pricing
Pricing model: Freemium
Usage-based pricing with included hours on every plan. Free tier available with limited usage. Paid plans include Browser Sessions (thousands of concurrent instances), Browser Agents (>90% success rate), Browser Functions (serverless AWS Lambda for browser work), Agent Vaults (encrypted credential storage), Agent Identities (real emails/phones with OTP intercept), and Session Profiles (full browser state persistence). Contact sales or sign up at console.notte.cc for specific pricing tiers and limits.
Notte pros
- 1000+ concurrent browser instances
- <50ms low latency
- 99.9% uptime guarantee
- >90% agent success rate
- Model agnostic - works with any LLM
- 3 lines of code for easy setup
- Built-in CAPTCHA solving and anti-detection
- Residential proxies with global network and fixed IPs
- Enterprise-grade AES-256 encrypted credential vaults
- Digital personas with real emails and phone numbers
- Auto-intercept OTPs and pass 2FA on any platform
- Full browser state persistence across sessions
- Live view and session playback for debugging
- Autoscaling for instant concurrent scaling
- Hybrid workflows cut costs by 50%+
- Playwright-compatible primitives
- Structured output with Pydantic models
- Serverless browser functions with 0ms network hop
- <200ms cold start for functions
- Cron scheduling for automated workflows
- Works with existing stack (Playwright, Puppeteer, Selenium)
- No selector maintenance needed for agents
- Self-healing workflows that don't break
- Open-source core framework
Notte cons
- Requires API key authentication
- CLI-first workflow needed for reliable results
- Cannot hand-write selectors from memory - must observe live
- Interactive browser login required for setup
- Learning curve for CLI commands
- SDK docs only for editing exported code not starting fresh
- Headless mode requires viewer URL to watch sessions
- Guessed selectors fail on real pages
- May need residential proxies to avoid bot detection
- Session configuration cannot change mid-session
- Open-source core requires own LLM API keys
- Pricing details not prominently displayed on homepage
- Console access required for full management
- Function deployment needed for reusable automations
- Rate limits apply by plan tier
Frequently asked questions about Notte
What is Notte?
Notte is a full-stack framework for building and running AI-powered browser agents that automate complex, multi-step workflows on the web. It provides browser infrastructure that lets AI run on the internet at speed, with cloud browser sessions, web agents, scraping, serverless functions, credentials vaults, and digital identities all through a single API.
How do I get started with Notte?
Install the Notte CLI using Homebrew (brew tap nottelabs/notte-cli && brew install notte) or Go (go install github.com/nottelabs/notte-cli/cmd/notte@latest), then authenticate with 'notte auth login' which opens a browser login flow. After authentication, start a session with 'notte sessions start', navigate to your target page, observe it, and execute actions. Export workflow code with 'notte sessions workflow-code'.
What makes Notte agents reliable?
Notte uses a hybrid architecture that combines deterministic scripts with AI reasoning. You script deterministic parts and use AI only when needed, cutting costs by 50%+ while improving reliability. Agents have >90% success rate, built-in anti-detection, CAPTCHA solving, and self-healing workflows that don't break when selectors change.
Does Notte work with my existing tools?
Yes, Notte is compatible with Playwright, Puppeteer, Selenium for browser automation; Python, TypeScript, Node.js for languages; Docker for containerization; and LLM providers including OpenAI, Anthropic, and Gemini. It also integrates with LangChain, Claude Code, Cursor, Stagehand, and browser-use frameworks. Drop in one line of code with no rewrites or lock-in.
How do I handle authentication and credentials?
Notte provides Agent Vaults with AES-256 encryption at rest, scoped access per session, automatic rotation, and full audit logs. Agents fetch secrets at runtime, never hardcoded. You can also use Agent Identities with real emails and phone numbers to receive OTPs, pass 2FA, and verify accounts on any platform. Session Profiles save full browser state so you log in once and skip auth on future runs.
What is the Anything API?
The Anything API lets you describe any web workflow task and Notte engineers it as a clean API endpoint. Your agent calls it with one POST request to api.notte.cc/functions/run. No agent improvising - Notte handles the engineering. For example, describe 'Get listing details for any Airbnb URL' and get back structured JSON with title, price, rating, reviews, and superhost status.
How do Browser Functions work?
Browser Functions are AWS Lambda for browser work - they let you colocation automations and agents close to browsers with 0ms network hop and <200ms cold start. Create functions from workflow files, invoke them via HTTP API/SDK/cURL/webhooks, schedule them with cron expressions, and manage versions. Functions are the deployed endpoint form of tested browser tasks.
Can I avoid CAPTCHAs and bot detection?
Yes, Notte sessions include built-in CAPTCHA solving, residential proxies with global network and fixed IPs, stealth mode with anti-detection features, ad blocking, tracker blocking, and browser profiles. If you get blocked, start a session with --proxy flag to use residential proxies. The Anti-Detection Playbook in docs provides additional guidance.
What is the difference between CLI and SDK?
The CLI is the source of truth for real browser state - use it to inspect live sites, validate actions/selectors, handle auth/session state, and export workflow code. The SDK is for understanding, running, or editing exported workflow code only. SDK-first automation is guesswork on real pages with dynamic selectors, auth state, CAPTCHAs, and anti-bot behavior. Always start with CLI:
How do I scrape data with Notte?
Use the scraping endpoint for fast data extraction - it automatically creates and manages sessions. Pass custom instructions for structured outputs and enable stealth mode. For structured scraping, define a Pydantic model and pass it as response_format. Example: cli.scrape(url='https://example.com', scrape_links=True, only_main_content=True) or with custom instructions to extract specific fields like title, date, and content.