VulpineOS

VulpineOS is an open-source AI browser agent runtime that gives autonomous agents a hardened, self-hostable browser to run reliable web interactions. Built on Camoufox (a fingerprint-hardened Firefox 146 fork), it acts as the identity and execution layer for AI browser agents, pairing a stealth-hardened browser session with operator tooling so agents stop breaking on selector failures and site defenses.

Its security stack includes prompt-injection filtering, an action-lock/page-freeze that halts a session mid-run, and full audit trails for every agent action. For data work, VulpineOS ships token-optimized DOM export that strips pages down to structured, LLM-friendly output to cut token costs on extraction. Teams can run multi-agent orchestration with budgets and handoffs, and self-host the entire stack via Docker with custom proxy pools and internal infrastructure.

The runtime is aimed at three groups: agent builders tired of brittle selectors, data teams needing reproducible extraction at lower token cost, and infrastructure owners who want browser automation running inside their own environment. Released under the MPL 2.0 license with code on GitHub, VulpineOS is currently pre-launch behind a waitlist, offering early-access credits, custom guides, and direct team access. It is a fit for developers building production browser agents who need injection filtering, auditability, and full control over where the browser runs.

Developers and data teams use VulpineOS to run reliable, self-hosted AI browser agents with injection filtering, auditability, and low-token DOM extraction.