$ loading_
Let AI control a real browser for navigation, forms, and web tasks.
This tool is described as exposing Playwright browser automation as MCP tools, allowing control of a real browser while preserving the user's active session; that implies meaningful local execution and session-data access, but the materials do not show any required secrets or fixed remote egress endpoints. Open source is a positive factor, but the repo has low adoption, no declared license, and unknown maintenance, so the overall posture is caution.
The materials explicitly state that no keys or environment variables are required, and there is no indication of API tokens, account passwords, or other long-lived secrets being requested. The main concern shifts to automated use of existing browser sessions rather than collection or storage of new credentials.
No dedicated remote API endpoint is declared for the tool itself, but its purpose is to control a real browser for navigation and form filling, so normal network traffic to visited websites should be expected and user inputs may be submitted to those sites. The materials do not show additional data exfiltration to unrelated or unknown third-party services.
The system flags it as executes-code, and the description indicates local browser automation via Playwright, which is a typical local process-launching and automation capability. The current materials do not show requests for system privileges clearly beyond its stated function.
The description emphasizes 'preserving the user's active session,' which implies the tool can read page content, manipulate forms, and act within already authenticated websites on the user's behalf. That creates broad access to session-linked web data and actions. The materials do not state direct arbitrary local file read/write access, so there is no clear evidence of over-privileged data access from the provided facts.
A positive factor is that the repository is open source and therefore auditable. However, the source is only a third-party registry entry, the GitHub repo has 0 stars, no declared license, and unknown maintenance status, which means low community validation. This is not enough on its own to rate it as high risk, but source and dependency review is advisable before use.
Copy the install command and let the AI configure it · recommended for beginners
No copy-paste install info for "mcp-browser" yet — see the docs or source repo.
Open the application page on the job website, fill in my name, email, phone number, and resume details, then pause for my confirmation before submitting.
The AI fills out the form in a real browser and waits for confirmation before final submission.
Use the browser to open my staging site, verify login, navigation, and search work correctly, and record the result of each step with screenshots of any issues.
You get a web testing report with steps performed, results, and screenshots of any issues.
Open these e-commerce product pages, extract the product name, price, rating, and stock status, and organize them into a table.
It outputs a structured table summarizing key information from multiple web pages.
Automate browser tasks, capture console logs, and take screenshots for web workflows.
Let AI control a browser for web tasks, extraction, and testing.
Control a browser with AI for automation, extraction, interception, and screenshots.
Let AI control browsers for search, interaction, extraction, and tab automation.
Let AI browse via your real Chrome for extraction and multi-step workflows.
Control a real local browser for web automation, extraction, and screenshots.