Skip to content

Core concepts ​

Four objects explain almost everything in BrowserPilot. The rest of the docs assume them.

Environment ​

An environment is one isolated browser identity: its own window, its own cookies, its own storage, its own settings for permissions.

  • Many environments can run in one process; each has one window.
  • Closing a window keeps the environment's logins. Deleting the environment wipes them.
  • The default environment is created on first launch and cannot be deleted.

Environments are how "log in once, let the agent work" stays safe: the agent only ever sees the environment you point it at.

Tab ​

A tab is a page inside an environment window. Tabs have globally unique ids; tools accept either env (act on the active page of that environment) or a precise tabId.

Every tab has a control state:

StateMeaning
userThe human is driving; the agent's input is not applied
agentThe agent is driving; an overlay covers the page
sharedBoth may act; used after a handoff

page_handoff releases control to the human, page_unlock removes the overlay without pausing the task, page_lock takes control back.

Run ​

A run is one execution of a workflow. It records the start time, whether it succeeded, the node that failed, the error message, declared outputs and a failure screenshot.

Runs are listed in the in-app Runs page; agents read them through the workflow tools.

Step ​

A step is a node inside a workflow graph. Node types cover navigation, clicking, typing, keys, scrolling, selecting, uploads, canvas strokes, waiting, branching, loops, page scripts, main-process code, raw CDP, HTTP requests, extraction and CSV export.

Edges connect steps. A condition node exposes one output per branch; a loop node contains its own subgraph.

How a task flows ​

text
you describe / a workflow is chosen
        ↓
agent picks tools (MCP) or workflow_run
        ↓
steps execute in one tab, with waits and retries
        ↓
outputs, files and a run record

Where things are stored ​

Everything lives in the app data directory: settings, environments, workflows, sessions and the MCP token. Cookies sit inside each environment's session partition. Nothing is uploaded anywhere; the app only downloads.

Next ​

Source is public; no LICENSE in the repository yet · GitHub · Releases · Issues