Quickstart
1
The Browserbase Agents page
Open the Browserbase Dashboard and navigate to the Agents page:

2
Create your first Browserbase Agent
Create a “Job Finder” Agent that, given a job title and a company name, returns the best fit open position.Click on the top right “Create Agent” button, and provide the following The above prompt defines the Agent’s goal and scopes its work by defining high-level steps and success definition.Now, instruct the desired structured output with a JSON Schema:Your “Job Finder” form should look as follows:
Click on “Create Agent”
system prompt:
3
Trigger your first Agent run
Creating an Agent redirects to its page where you can:
Provide your own prompt or use the following one:Then, click on “Run”.
- edit its config
- quickly access key statistics (success rate, average duration)
- see the list of runs

4
Your first Agent run
Triggering an Agent run redirects to the run view where you can inspect:
- every step of the run: reasoning, actions, and more
- see the live browser session
- interact with the timeline to inspect a specific step

5
Optimize your Agent
The first Agent run is always exploratory. The Agent could face issues for the first time, resulting in slow, repeated steps.Browserbase Agents provide an “Optimize” feature, available from the Agent run view, that suggests optimization based on your
Agent’s previous exploration.From the Agent run view, click on the “Optimize” button at the top right and select “Make it faster”:
No matter what your selected prompt was, the Optimize tool will likely suggest updating your Agent’s prompt to always navigate to an Ashby page if available (instead of exploring the company website).By accepting the suggested solution, the Optimize tool will update your Agent’s prompt and trigger a new run:
By updating the prompt, your Agent should now run 30% faster!



When should you use the Optimize tool?
- If you want to speed up your Agent
- If your Agent success rate is dropping, ask it to find solutions
Next steps
Agents overview
Learn run states, configuration options, observability, and production notes.
How it works
The execution loop, built-in tools, run lifecycle, and session observability
Node.js SDK
Create an Agent, start runs, and read results from Node.js.
Python SDK
Create an Agent, start runs, and read results from Python.
Agents API reference
See the request and response fields for creating a run.
Browser agents use case
Understand when to use Agents versus a custom browser-agent stack.