Base44 Superagent Gains Browser Control and Live Voice Conversations
Base44Base44 gave Superagent two major new interaction modes on the same day: the ability to control a user's own Chrome browser via a dedicated extension (clicking, navigating, and reading pages directly rather than using a separate cloud-hosted browser), and the ability to hold live voice conversations directly in the editor. Together, the additions meaningfully expand how directly and naturally users can work with their Superagent.
Two New Ways to Work With Your Superagent
Base44 shipped a pair of significant Superagent capability upgrades on the same release day, both aimed at making the agent feel less like a chat window and more like a genuine collaborator.
Browser Control From Your Own Chrome
With the Superagent by Base44 Chrome extension installed, Superagents can now click, navigate, and read pages directly inside a user's own browser session, rather than relying on a separate cloud-hosted browser instance. This is a meaningful shift: previously, agent-driven browsing happened in an isolated environment disconnected from the user's actual logged-in sessions, cookies, and context. Now the agent can act within the browser tabs a person is already using, useful for tasks like filling out forms on sites the user is authenticated into, or verifying something live on a page without switching context. Browser actions carry small per-action credit costs (roughly 0.01 credits for clicks and reads, 0.03 for data extraction), with no charge for failed attempts.
Talking to Your Superagent
Base44 also introduced live voice conversations with Superagent, directly in the editor. Rather than typing or dictating single messages, users can now have a real-time back-and-forth spoken exchange with their agent, a step up from the speech-to-text dictation feature Base44 shipped for mobile Superagent earlier in July.
Why It Matters
Both features push Superagent further from a text-based assistant toward something that operates more like a real digital coworker: one that can see and act on the same web pages a person is looking at, and one that can be talked to naturally rather than typed at. For builders managing multiple tasks at once, voice conversation in particular lowers the friction of interacting with an agent while doing other work.