At a glance
| Item | Value |
|---|---|
| Availability | Self-serve[1] |
| Billing unit | Dollar-denominated credits spent per hour of browser session ($0.01 per hour); the page says usage is rounded by a billing policy without stating the increment[1] |
| Free tier | $10 credits granted once with no expiry; 5 concurrent browsers; 15-minute sessions; no card required[1] |
| Entry paid plan | Developer ($20/mo)[1] |
| Browser-hour rate (self-serve plans) | $0.01[1] |
| Max concurrency | Browsers: 5 Free, 25 Developer, 100 Startup, Custom Enterprise. Agents and functions: 5, 25, 100, Custom.[1] |
| Max session duration | Per plan: 15 min Free, 120 min Developer, 300 min Startup, custom on Enterprise. The API parameter max_duration_minutes documents a schema maximum of 1440 minutes, which exceeds the self-serve plan limits.[1] |
| Idle timeout | Default 3 minutes, configurable from 1 to 30 minutes; each operation resets it[2] |
| Regions | Not published |
| SDK languages | python, typescript[3] |
| Open source | Yes: SSPL-1.0 (Server Side Public License v1), github.com/nottelabs/notte. The core library (local browser sessions and scraping) only; the hosted API service (stealth sessions, functions, vaults and personas) is not open source.[3] |
| SOC 2 | Yes: Trust Center lists a SOC2 type 2 report.[4] |
| HIPAA | In progress: Trust Center lists HIPAA as In progress.[4] |
| GDPR and DPA | In progress: Trust Center lists GDPR as In progress; no DPA statement was found in the pages read.[4] |
Sources: notte.cc [1] ↗docs.notte.cc [2] ↗github.com [3] ↗trust.notte.cc [4] ↗
Pricing
| Plan | Monthly | Included | Browser-hour rate | Concurrency | Max session | Proxy per GB | Retention |
|---|---|---|---|---|---|---|---|
| Free[1] | $0 | Not published | $0.01 | 5 | 15 min | $10 | Not published |
| Developer[1] | $20/mo | Not published | $0.01 | 25 | 2 hours | $10 | Not published |
| Startup[1] | $100/mo | Not published | $0.01 | 100 | 5 hours | $10 | Not published |
| Enterprise[1] | Not published | Not published | Not published | Not published | Not published | Not published | Not published |
- Free, as published: $0.01 / hour of browser sessions, paid from credits ($10 credits, granted once)[1]
- Developer, as published: $20/mo including $20 credits per month; usage then $0.01 / hour of browser sessions from credits[1]
- Startup, as published: $100/mo including $100 credits per month; usage then $0.01 / hour of browser sessions from credits[1]
- Enterprise, as published: Custom[1]
Sources: notte.cc ↗
Proxy, CAPTCHA and other meters
The notes below were recorded with the sourced values. They are not quotations from the provider.
| Meter | Pricing |
|---|---|
| Proxy | Built-in residential proxies $10 per GB. Custom proxies on the Startup plan $0.20 per GB, on top of what the proxy provider charges. Bandwidth is counted as request plus response bytes, rounded up to the next MB per session.[1] |
| CAPTCHA solving | Included on Free, Developer and Startup (no per-solve charge listed); Enterprise row says Targeted[2] |
| Browser sessions | $0.01 per hour, rounded per billing policy[2] |
| Browser functions | $0.05 per hour, rounded per billing policy[2] |
| Residential proxies | $10 per GB[2] |
| LLM tokens (agents, scraping extraction) | Token price; the public pricing page says no Notte markup when using Notte-managed keys (see notes for a conflicting docs page)[2] |
| Personas (email and phone) | Additional charges for the email and phone services; amounts not stated on the page[3] |
| Manual credit top-ups | Available; purchased credits do not expire[2] |
| Session file retention (per plan) | Free 1 day, Developer 7 days, Startup 30 days, Enterprise custom[4] |
- Everything is paid from dollar credits. Monthly bill estimate: plan fee + max(0, usage - plan credits), where usage = browser hours x $0.01 + function hours x $0.05 + proxy GB x $10 (or BYO proxy GB x $0.20 on Startup) + LLM tokens + persona fees. Monthly plan credits expire each cycle; free credits and top-ups do not.
- The public pricing page does not state what happens after credits are exhausted on Free; it lists manual top-ups as available.
- The pricing page does not state the browser-time rounding increment ('Rounded by billing policy'). The docs llms-full.txt copy of a Pricing page says 'Charged per minute, rounded up.' but docs.notte.cc/pricing now redirects to www.notte.cc/pricing, so that text may be stale.
- Advanced stealth (highest-fidelity browser environment) is available to approved workspaces on request, per the session configuration docs; the pricing page lists captcha solving and stealth as Included for self-serve plans and Targeted for Enterprise.
Sources: docs.notte.cc [1] ↗notte.cc [2] ↗docs.notte.cc [3] ↗docs.notte.cc [4] ↗
Limits
| Limit | Value |
|---|---|
| Max session duration | Per plan: 15 min Free, 120 min Developer, 300 min Startup, custom on Enterprise. The API parameter max_duration_minutes documents a schema maximum of 1440 minutes, which exceeds the self-serve plan limits.[1] |
| Default session timeout | 15 minutes (max_duration_minutes default)[2] |
| Idle timeout | Default 3 minutes, configurable from 1 to 30 minutes; each operation resets it[2] |
| Max concurrency | Browsers: 5 Free, 25 Developer, 100 Startup, Custom Enterprise. Agents and functions: 5, 25, 100, Custom.[1] |
| Session creation rate limit | Rate limited per API key with a separate budget for session starts (POST /sessions/start): per minute 50 (Free), 50 (Developer), 100 (Startup); per hour 200, 200, 1,000; per day 600, 600, 10,000; Enterprise custom. Requests over the limit return 429.[3] |
| Regions | Not published |
Sources: notte.cc [1] ↗docs.notte.cc [2] ↗docs.notte.cc [3] ↗
Features
| Feature | Group | Status | Detail |
|---|---|---|---|
| Connect over CDP | Connect | Yes[1] | Every session exposes a CDP endpoint (cdp_url) that Playwright connects to with chromium.connect_over_cdp() in Python or chromium.connectOverCDP() in Node. |
| Playwright | Connect | Yes[1] | Connect Playwright over CDP with connect_over_cdp; the SDK also exposes a session.page Playwright page. |
| Puppeteer | Connect | Yes[2] | Use puppeteer-core with the session CDP URL from the Node.js SDK, Python SDK or REST API. |
| Selenium | Connect | Yes[3] | Documented as Python and TypeScript examples in the notte repo that connect a local ChromeDriver to the hosted browser through a loopback bridge bound to 127.0.0.1; Chrome runs on Notte. ChromeDriver remote debugging restrictions apply, including unsupported window resizing. |
| Persistent profiles | Login and state | Yes[4] | Profiles snapshot cookies, local storage and session storage (not HTTP cache, history or extensions); attach with profile={id, persist} or save a running session with save_profile(). Default is read-only (persist=False). Profile API supports create, duplicate, list, get, delete, cookies get/set and state set. |
| Credential vault | Login and state | Yes[5] | Vaults store credentials such as email, password and MFA values; sessions fill placeholder values and the vault substitutes the real credentials outside the model context for supported fill actions. |
| Managed login | Login and state | Yes[6] | Managed Auth: create connections through a connect-link flow, then pass up to ten connection IDs in auth_ids when creating a session; Notte verifies the connections and logs in when needed. Expired credentials or connections needing user interaction require the connection flow. |
| TOTP and 2FA handling | Login and state | Yes[7] | Personas have their own email inbox and an SMS-capable phone number for verification and 2FA, and the docs list 2FA support for personas and for the vault (an MFA placeholder exists). TOTP authenticator seed handling is not described in the pages read. |
| CAPTCHA solving | Anti-bot and network | Yes[8] | On by default (solve_captchas=True): reCAPTCHA v2, hCaptcha, Cloudflare Turnstile, press-and-hold and simple text CAPTCHAs. The docs state it cannot solve 100% of captchas, heavily distorted text challenges or custom/proprietary systems. |
| Stealth and fingerprinting | Anti-bot and network | Yes[9] | Every session gets a coherent recent Chrome desktop fingerprint (user agent, platform, screen, graphics hardware), with timezone, locale and geolocation following the proxy exit IP. advanced_stealth=True runs the session in a higher-fidelity browser environment for approved workspaces. |
| Managed proxies | Anti-bot and network | Yes[10] | Built-in residential proxies are on by default (proxies=True) and exit in the US unless a country is chosen; each session uses a single proxy. Billed at $10 per GB. |
| Proxy geo-targeting | Anti-bot and network | Yes[10] | Country-level: pass a two-letter country code (for example proxies='fr'). The number of supported countries and city-level targeting are not stated in the pages read. |
| Bring your own proxy | Anti-bot and network | Yes[10] | Custom HTTP or HTTPS proxies, listed as a Startup plan feature and billed at $0.20 per GB of bandwidth on top of the proxy provider's charges. Enterprise plans can have custom terms. |
| Live view | Observe and control | Yes[11] | Browser, CDP and Jupyter notebook viewers; the shareable signed viewer_url stays valid for 12 hours after it is issued. |
| Embeddable live view | Observe and control | Yes[12] | viewer_url from the session status response can be rendered in an iframe; query parameters include mode=embed-minimal, interactive and theme. URL is JWT-signed to the session. |
| Human takeover | Observe and control | Yes[12] | The embedded viewer is interactive; the interactive=1 parameter requires the user to click the iframe once before controlling the browser, and interactive=0 disables the controls. |
| Session recording | Observe and control | Yes[13] | Recording is on by default (record_session=True, cannot be changed after creation); MP4 replay via session.replay() or GET /sessions/{id}/replay, presigned URL expires after 1 hour. Not recorded: audio, network request details, console logs. |
| Network logs | Observe and control | Yes[14] | log_session (default true) saves steps, raw CDP messages, network logs and API request/response content; network logs are read with GET /sessions/{id}/network/logs. No HAR format is mentioned. |
| File downloads | Files and extensions | Yes[15] | Session file storage: the download_file action and storage.download() retrieve files the browser saved; files are session-owned. |
| File uploads | Files and extensions | Yes[15] | storage.upload() then the upload_file action attaches a file to a form input. |
| Browser extensions | Files and extensions | No[16] | Notte-hosted sessions do not load custom Chrome extensions; the documented workaround is to launch the extension browser elsewhere and connect Notte to it with cdp_url. |
| Serverless functions | Run code and agents | Yes[17] | Python scripts with a run() function, deployed as Functions that run on Notte infrastructure and can be invoked via API or scheduled. A maximum function runtime is not stated in the pages read. |
| Scheduling | Run code and agents | Yes[18] | One schedule per Function, using the AWS EventBridge six-field cron format or rate() expressions, firing in UTC; managed in the Console, over HTTP or with the TypeScript SDK. The Python SDK has no schedule methods yet. |
| Webhooks | Run code and agents | Yes[19] | Functions can be invoked as inbound webhook endpoints (POST /functions/{function_id}/runs/start, authenticated with the API key). No outbound event webhook for session or function-run completion was found in the pages read. |
| MCP server | Run code and agents | Yes[20] | Hosted at https://api.notte.cc/mcp with OAuth sign-in or Notte API key bearer auth. A separate endpoint at /mcp/full exposes the broader developer API, including function and agent administration. |
| Hosted agent API | Run code and agents | Yes[21] | The pricing page lists agents among the core features available across plans, with agent LLM usage billed at token price and no Notte markup. |
| Scrape and fetch API | Run code and agents | Yes[22] | POST /scrape (SDK client.scrape) returns markdown or structured data with LLM extraction; a separate search endpoint returns URLs, snippets or a sourced answer. A per-session screenshot endpoint exists; no standalone PDF endpoint was found. |
| Self-hosting | Deployment | Enterprise[21] | On-premise deployment (cloud on-prem) is listed only for Enterprise. Separately the open-source repo (SSPL v1) runs local Playwright-based sessions and scraping with no API key, but not the hosted stealth, captcha and proxy stack. |
| Region selection | Deployment | Unknown | No region selection or region list found in the pages read. |
Sources (22)
docs.notte.cc [1] ↗docs.notte.cc [2] ↗docs.notte.cc [3] ↗docs.notte.cc [4] ↗docs.notte.cc [5] ↗docs.notte.cc [6] ↗docs.notte.cc [7] ↗docs.notte.cc [8] ↗docs.notte.cc [9] ↗docs.notte.cc [10] ↗docs.notte.cc [11] ↗docs.notte.cc [12] ↗docs.notte.cc [13] ↗docs.notte.cc [14] ↗docs.notte.cc [15] ↗docs.notte.cc [16] ↗docs.notte.cc [17] ↗docs.notte.cc [18] ↗docs.notte.cc [19] ↗docs.notte.cc [20] ↗notte.cc [21] ↗docs.notte.cc [22] ↗
Get started
SDK languages: python, typescript. The snippet below is the official python quickstart from docs.notte.cc, trimmed.
- python: notte-sdk[2]
- typescript: notte-sdk[3]
- python: notte (open-source core: local sessions and scraping)[4]
import os
from notte_sdk import NotteClient
from playwright.sync_api import sync_playwright
client = NotteClient()
with client.Session(idle_timeout_minutes=2) as session:
cdp_url = session.cdp_url()
with sync_playwright() as p:
browser = p.chromium.connect_over_cdp(cdp_url)
try:
page = browser.contexts[0].pages[0]
page.goto("https://www.google.com")
page.screenshot(path=os.environ.get("NOTTE_SCREENSHOT_PATH", "screenshot.png"))
finally:
browser.close()Sources: docs.notte.cc [1] ↗pypi.org [2] ↗npmjs.com [3] ↗pypi.org [4] ↗
Recent launches
- 2026-10-23: Python SDK 1.10.0: async client (preview), regenerated resource operations[1]
- 2026-10-22: v1.10: persona email_provider options (AgentMail, Resend); agent loop removed from the Python package per README[2]
- 2026-09-24: Hosted MCP server replaced with eight curated tools (switch made Sept 4, per blog post)[3]
- 2026-09-16: Browser payments with Link: connect wallet, approve a spend, fill checkout with a temporary card[4]
- 2026-09-15: Coordinated browser actions and CAPTCHA solving in the browser runtime[5]
- 2026-09-12: Blog: authorized-agent verification and CAPTCHAs (Web Bot Auth context)[6]
Sources: github.com [1] ↗github.com [2] ↗notte.cc [3] ↗notte.cc [4] ↗notte.cc [5] ↗notte.cc [6] ↗