Human-in-the-loop agent
Build a human-in-the-loop AI agent that pauses for user input mid-task, then resumes autonomously with Stagehand and Browserbase.
Send this template to your email, or view it on your desktop for the source code.
Get started with human-in-the-loop automation
Give a Stagehand Agent a task and let it work autonomously. When it hits something it can't answer, it pauses and asks the human via a custom `askHuman` tool, then picks up where it left off.
Watch the whole thing happen in real time through Browserbase Live View with an SSE-powered activity log streaming every step. This template wires up the full loop: agent execution, human input collection, and resume logic, all in a single Next.js app. Perfect for assisted form filling, approval workflows, and supervised data entry where some decisions need a human in the loop.
Steps
- Start a Browserbase session and navigate to the target job application page.
- The Stagehand Agent begins filling out the application autonomously.
- When the agent encounters a field it cannot answer, it invokes the askHuman tool to pause and prompt the user.
- The user responds via the frontend UI, and the agent resumes with the provided information.
- The agent completes and submits the application, streaming progress via SSE throughout.