A live browser inside every agent
Agents now drive a real Chromium session you can watch in real time, and take over whenever you like. It changes what an agent can verify.
An agent that cannot see the thing it built is guessing. It can read the code, run the tests, and still miss that the button renders behind the header. Four days after launch, every JackHamr agent has eyes.
A real browser, not a scraper
Inside each agent's machine there is now a full Chromium browser. The agent can navigate, click, type, scroll, and read the page exactly the way a person would. The session streams live into the dashboard, so you watch the cursor move as the agent works through a checkout flow or debugs the app it just deployed.
When you want to step in, you take the mouse and keyboard. Log in to something the agent should not have credentials for, click through a step it is stuck on, then hand control back. The agent picks up where you left off.
Side by side with the chat
The workspace shows chat and browser in a split view. Drag the divider to whatever ratio suits you and the layout is remembered in the URL, so a link to a session opens the way you left it.
- The browser starts automatically the first time an agent needs it
- Sessions keep their URL and can be restarted in place
- Zoom, pan, and an on-screen keyboard on mobile followed in April
What this unlocks
This is the release that makes "build it and test it" honest. An agent can now ship a feature, open the running app, walk through the user flow, and tell you what it saw. Later, when QA agents and reviewer agents arrived, this is the browser they use to try to break the builder's work before it reaches you.
Agents that can look are agents you can trust with more.
More releases
Autonomous agents: give it a job, not a prompt
Proactive agents that work toward your goals on their own. Give one a job, sign off on its plan, and it checks in, works through its backlog, asks for approval where you want it, and gets better every week.
2 min read
Agents can talk to other agents
Approve trust between two agents and they can message each other, or mention a trusted agent into your own chat. A fleet that coordinates, with guardrails.
2 min read
Agents that learn: the memory wiki
Every agent learns on its own. Each session ends with notes, a nightly librarian merges them into a wiki, and the wiki shapes every prompt after that. Month three looks nothing like day one.
2 min read