Hyper brings cloud browsing to its stable app
Hyper 1.11.0 brings cloud browsing to its stable release, according to the company’s October 17 notes. The assistant can navigate websites, type and manage multiple tabs.
The company’s October 9 release had described cloud browsing as available in nightly builds but disabled in stable. That history makes this an expansion of availability. People following the regular release channel can now try a capability previously reserved for preview builds.
An extra route into workplace software
Hyper presents itself as a workplace AI app built around company knowledge. Its documentation describes drawing context from messages, documents, meetings, email and code so teams can research questions or prepare work without assembling that background for each request.
The October 9 release had already expanded its connector catalog to services including Salesforce, Figma and Shopify. It also introduced parallel assistants and more cards for reviewing proposed actions. Those integrations have their own service permissions and supported operations.
Desktop control and temporary sessions
On desktop, users can watch a live preview and take control, pausing the agent. Users enter passwords and verification codes themselves. iPhone can display or skip these requests, but completing them requires desktop.
Each conversation has its own browser, separate from personal cookies. Sessions close after ten minutes without browser activity. A pending user request keeps the browser open until answered or expired. Sign ins disappear on closure and do not transfer between conversations. Some sites block cloud browsers, and browser approvals remain separate from connector permissions.
Hyper’s download page offers versions for Apple silicon Macs, Intel Macs and Windows PCs. ByteForward did not install the software or test an authenticated browser session.
Check the workflow before relying on it
The company’s pricing information advertises a three day trial and paid Sprint and Blitz plans with monthly token allowances. That makes usage relevant to an evaluation. A feature being available does not establish how much a particular sequence of browser actions will consume.
A useful pilot would start with a public page and a task that only reads information. Ask for a specific result, such as finding inconsistent product details across two pages. Check the result against both originals and record any correction needed before allowing changes to a real account. This keeps the first test easy to inspect.
For a second trial, define one reversible edit in a test account and check the saved state independently. Record the starting values and an undo step before the edit so a mistake has a clear recovery path. Measure completion time, manual interventions and failed attempts alongside the result. Those observations would make a stronger purchasing case than a successful demonstration alone.
The same need for inspectable results appears in ByteForward’s coverage of Cursor cloud agent canvases, which examines how teams can review and share an agent’s finished work.
Illustrative browser workspace with Chrome on a laptop. Photo by Startup Stock Photos via Pexels under CC0. Archival photograph from February 2015.
