Notte
Web-agent framework: scripted browsing + AI agents with stealth cloud sessions.
About
Notte is a full-stack framework for building AI agents that browse the web. Its pitch is hybrid: script the deterministic parts of a workflow and spend AI tokens only where judgment is needed, which the vendor claims cuts costs by 50%+ versus pure-agent approaches. The open-core SDK covers web agents, structured Pydantic output, and Playwright-compatible primitives; the hosted API adds stealth browser sessions (CAPTCHA solving, proxies, anti-detection), secret vaults, and digital personas. An MCP package (notte-mcp) exposes it to assistants. Note the license: SSPL-1.0 — source-available, not OSI open source, despite 'open source' marketing language.
What it can do
- **Hybrid workflows**: mix deterministic scripting with AI agents in one flow
- **Stealth sessions**: cloud browsers with CAPTCHA solving, proxies, anti-detection
- **Structured output**: Pydantic models for scraped data
- **Secrets vaults**: credential storage (emails, passwords, MFA, SSO) for agents
- **Digital personas**: generated identities for account-creation workflows
- **MCP package**: `notte-mcp` connects web agents to Claude/Cursor-class assistants
Setup
```bash
pip install notte
patchright install --with-deps chromium
```
```python
import notte
with notte.Session(headless=False) as session:
agent = notte.Agent(session=session, max_steps=10)
response = agent.run(task="Find three cat memes on Google Images")
```
Local mode needs your own LLM API keys. Hosted mode (stealth sessions, vaults) needs an API key from console.notte.cc. Node SDK: npm install notte-sdk. Docs: docs.notte.cc.
Prerequisites
Python 3.11+ (or Node.js for the TS SDK); Chromium via patchright. LLM API keys for local mode; a Notte API key for hosted sessions.
Use cases
Works with
Trust score — 7/10
Scored 7/10 (metrics tier, metrics-v1). Evidence: 2,026 GitHub stars (log-scaled vs catalog peers); blended with ~1,766 monthly package downloads; last push 0d ago; 62 releases/12mo; 11 open issues. Manual review: hybrid script+AI model is genuinely useful; only 2k stars / 1.7k PyPI-mo; benchmark claims are vendor-run; hosted path touches vendor infra with real credentials. License: SSPL-1.0 — source-available, NOT OSI open source (despite 'open source' marketing).
★ 2,026 GitHub stars
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 6.1/10
- Maintenance · recency, releases, issue load
- 9.8/10
- Openness · license + docs
- 7.2/10
- Manual review · sanity read + red-flag check
- 7.0/10
- · 2,026 GitHub stars (log-scaled vs catalog peers)
- · blended with ~1,766 monthly package downloads
- · last push 0d ago; 62 releases/12mo; 11 open issues
- · manual review: hybrid script+AI model is genuinely useful; only 2k stars / 1.7k PyPI-mo; benchmark claims are vendor-run; hosted path touches vendor infra with real credentials
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.