SHIPPED · v1.3.0
Browser profiles, seven more coding agents and media uploads over MCP
New
Browser profiles. Settings → Profiles holds one profile per identity, like Founder, Support or a client account: a name, a badge emoji or initials, the main email, notes and the sites it is signed in to, each with a Sign in button. "Sign out of all sites" and "Show data folder" are on the profile too.
Pick a workspace's profile from its right-click menu: Browser profile lists your profiles and "Own session". The workspace switches right away and its open tabs reload signed in as that profile. Several workspaces can share one profile, and a workspace that leaves a profile gets its own cookies back.
A proxy or VPN per profile. Turn on "Send this profile's traffic through a proxy" and enter a SOCKS5 or HTTP proxy, such as the one your VPN provider gives you. Only that profile's workspaces use it. Proxies that ask for a user name and password are not supported yet.
Import several browser profiles at once. Settings → Import Profile lets you tick profiles, or all, and imports them one after another. By default each one gets a new workspace named after it, so separate sign-ins stay separate; you can also send them all to one workspace. Profiles that fail stay ticked so you can retry them.
Seven more coding agents in chat: Gemini CLI (with Auto, Pro, Flash and Flash-Lite models), GitHub Copilot CLI, Factory Droid, Augment (Auggie), Kilo Code, Qoder and Command Code.
Upload media over MCP access. browser_upload_file now takes public image and video URLs as well as files on your computer, so an agent can attach media to a post on YouTube, a social network or any site with an upload button. Addresses on your computer or private network are refused.
New browser_drop_files tool, for drop zones and post composers that take files by drag and drop or paste instead of an upload button.
New browser_upload_status tool, which reads an upload's progress from the page (uploading, processing, complete or failed) and can wait until it finishes, so the agent only posts once the media is ready.
Improved
Coding agents show their own logos in the agent picker, in Ask AI and in the setup steps, instead of a generic icon.
browser_click can click a point on the page by its x and y coordinates.
Fixed
Fixed agents that stayed paused for good after asking you to sign in. If the agent stopped or its connection closed while waiting, every later browser tool was refused as "paused while the owner signs in". Now the pause ends too.

Settings → Profiles with Founder, Support and a client profile, and the Founder profile’s accounts and proxy 
The workspace right-click menu with Browser profile open, listing Own session and three profiles 
Settings → Import Profile with three browser profiles ticked and “A new workspace for each” chosen 
The Which app should help? picker listing Claude Code, Gemini CLI, GitHub Copilot CLI and other coding agents