LIVE
Google Introduces "Gemini 4 Carbon" as Next Step in Its Model Lineup28/10/26 · Google|DeepMind unveils Genie 4, a new generation of world model27/10/26 · Google DeepMind|Anthropic files IPO prospectus, with risk disclosures taking center stage26/10/26 · Anthropic|Robot policy trained only in simulation matches top real-world systems26/10/26|OpenAI unveils GPT-6.1 Terra, a model built for computer use26/10/26 · OpenAI|Google's AI gives evasive answers about who won the 2020 US election28/10/26 · Google|NVIDIA introduces Nemotron 3.5 Transcribe, a promptable multilingual transcription model28/10/26 · NVIDIA|OpenAI and ServiceNow deepen partnership on enterprise workflows28/10/26 · OpenAI|Failed startups' data finds a second life training AI models28/10/26|Anthropic expands AI for Science program to 20 foundations and research institutions28/10/26 · Anthropic|Anthropic shares early results on using Claude to accelerate drug development27/10/26 · Anthropic|GEPA-Curator: Filtering Failures to Better Train LLM Agents27/10/26|Google Introduces "Gemini 4 Carbon" as Next Step in Its Model Lineup28/10/26 · Google|DeepMind unveils Genie 4, a new generation of world model27/10/26 · Google DeepMind|Anthropic files IPO prospectus, with risk disclosures taking center stage26/10/26 · Anthropic|Robot policy trained only in simulation matches top real-world systems26/10/26|OpenAI unveils GPT-6.1 Terra, a model built for computer use26/10/26 · OpenAI|Google's AI gives evasive answers about who won the 2020 US election28/10/26 · Google|NVIDIA introduces Nemotron 3.5 Transcribe, a promptable multilingual transcription model28/10/26 · NVIDIA|OpenAI and ServiceNow deepen partnership on enterprise workflows28/10/26 · OpenAI|Failed startups' data finds a second life training AI models28/10/26|Anthropic expands AI for Science program to 20 foundations and research institutions28/10/26 · Anthropic|Anthropic shares early results on using Claude to accelerate drug development27/10/26 · Anthropic|GEPA-Curator: Filtering Failures to Better Train LLM Agents27/10/26|
ToolRisingofficial publisher

Browserbase et Stagehand

Tool : A precise action the AI can call · requires browserbase

Hosted Chrome browsers and act, extract, observe SDK for agents.

The most serious browser layer in production; a driven browser is an attack surface.

Overall score
6.7 /10
Security: serious reservations
Utility
8.0
Reliability
10.0
Security
Serious reservations
Ease
Technical · Technical

Security reflects where data goes, requested permissions and incidents on record at the verification date. It is not a guarantee of zero risk: check the deployment mode and the connections you enable.

What it does, concretely

Automating web tasks with an AI agent requires a reliable browser that can be controlled remotely and handle complex pages. Browserbase provides cloud-hosted Chrome browsers accessible via an API key, while Stagehand is the SDK that lets an agent drive them using natural language.

In practice, Stagehand exposes three primitives: act() to perform an action on the page (click, fill a form), observe() to identify what is interactive, and extract() to retrieve structured data. These methods rely on a language model of your choice and coexist with standard Playwright APIs for deterministic actions. When a site changes its structure, Stagehand detects the break and automatically refreshes how the action is executed.

The platform targets developers building web agents in TypeScript, Python, or Go who need production-ready browser infrastructure: isolated sessions, support for nested iframes, closed Shadow DOMs, OpenTelemetry observability, and search or page fetching without launching a full browser.

It is not suited for people without development skills, as there is no consumer-facing graphical interface. Furthermore, any agent that navigates, clicks, and fills forms acts on the real web, which requires careful attention to the permissions granted and the sites targeted.

What it offers

  • ·Cloud-hosted Chrome browsers accessible via a single API key for all sessions.
  • ·Three natural language primitives: act() to interact, observe() to analyse, extract() to retrieve structured data.
  • ·Self-healing actions: when a site's structure changes, Stagehand detects the break and automatically updates the selector used.
  • ·Compatibility with standard Playwright APIs for deterministic actions, alongside the AI primitives.
  • ·Native support for out-of-process nested iframes and closed Shadow DOMs, often unsupported by other tools.
  • ·Web search and page retrieval as Markdown via browserbase.search() and browserbase.fetch(), without launching a full browser.
  • ·Built-in OpenTelemetry observability to monitor agent actions in production.
  • ·SDK available in TypeScript, Python, and Go from a single monorepo.

In practice

Price
Freemium · from 18 €
Commercial use
Commercial use: yes
Available on
sdk-ts, sdk-python, mcp
Host
browserbase
FR interface
No

Security and data

Rating
Security: serious reservations
Capabilities
readswritesexecutesnetwork
Your data
Data with third parties
Maintenance
Active
Last release
23 October 2026

Links and repository checked on 25 October 2026 · added on 1 June 2024

How to install Browserbase et Stagehand

1. Create an account at browserbase.com and retrieve your API key from the dashboard. 2. Install the TypeScript SDK: npm install @browserbasehq/stagehand (or the Python/Go equivalent for your environment). 3. Export your keys as environment variables: export BROWSERBASE_API_KEY="..." and your LLM provider key. 4. Copy the quickstart example from docs.stagehand.dev and adapt the LLM model to your provider. 5. Precaution: only give the agent access to the sites and accounts strictly required; test on staging environments before any production use.
npm install @browserbasehq/stagehand

Official documentation ↗

In the same category

Whole category Automation & agents →