The workflow is to run the Auto Browser stack locally, connect an MCP client or the bundled stdio bridge, create a browser session, let a human handle sensitive login or recovery moments through noVNC when needed, save reusable auth profiles, and continue the task with screenshots, DOM summaries, downloads, audit events, approvals, and witness receipts available to the agent/operator.

This is not a generic browser automation listing. The boundary is supervised, local-first browser control for authorized MCP workflows: login-once profile reuse, human takeover, approval gates, audit trails, and repeatable browser task execution. It is a good fit for internal dashboards, admin tools, QA/debugging, and account workflows where an agent needs a live browser but the operator still needs control.

Inputs and prerequisites: Docker Compose, Auto Browser, MCP-compatible client or bundled stdio bridge, Playwright browser stack, optional noVNC operator dashboard.

Setup notes: Clone https://github.com/LvcidPsyche/auto-browser, run docker compose up –build, then connect an MCP client to the HTTP or stdio bridge. Optional Python clients are published as auto-browser-client and auto-browser-langchain.

Source and verification boundary: use https://github.com/LvcidPsyche/auto-browser as the canonical reference before running the workflow; keep commands, API calls, CLI usage, and generated outputs reviewable against that upstream source.

Framework fit: publish this as a MCP workflow only when the operator can invoke the documented toolchain directly, rather than treating the upstream project as a generic product listing.