browser-use 1.0.7 — browser-use 1.0.7
- Source
- github.com
- Author
- sarath-menonTop Viber
- Date

Context
Browser Use, an open-source project in which an AI model completes tasks in a real web browser by writing Python, fixed a timeout in version 1.0.7 that hit people driving it from coding tools such as Cursor (release notes). Its server, the interface that lets a coding assistant call Browser Use as a tool, offers a tool called browser_exec that runs a cell of code against the browser. According to the pull request behind the fix, clients such as Cursor cancel a tool call after 60 seconds. Before 1.0.7, a slow cell kept running, but the client gave up and the model received a timeout error with none of the output the cell had already produced.
In 1.0.7, browser_exec waits at most 50 seconds, leaving a margin under that limit, then replies with the output printed so far and a note that the code is still running. The model can call browser_exec again with empty code to wait for the cell and collect the rest of its output and how it ended. Calls that finish within 50 seconds reply exactly as before. A long browser task therefore no longer reaches the model as a failure that hides finished work: it sees partial progress and has a way to collect the result. The pull request also says that new code sent while a cell is still running runs alongside it, so the empty call is the way to pick up the remainder.
- MCP — The Model Context Protocol — an open standard that lets any AI assistant plug into any tool or data source without custom integration code.
Checking sign-in…
Loading comments…




