Acceptable Use Policy

Effective August 16, 2026

This policy is incorporated into the Terms of Service by reference. It exists because AgentBrowser gives automation/agents real, capable browser access — the abuse surface of that capability is concrete, not hypothetical, and this policy names it concretely rather than in vague boilerplate.

1. You must be authorized to do what you're automating

Use AgentBrowser only against systems, accounts, and data you're authorized to access. Automating a login, a workflow, or a data extraction you would not be allowed to do manually is not made permissible by doing it through AgentBrowser. You remain responsible for complying with the terms of service of any third-party site or API your automation interacts with.

2. Specifically prohibited

• Credential stuffing / account-takeover attempts. Using AgentBrowser's browser automation or vault credential-injection to attempt logins with credentials you don't have authorization to use, including testing leaked/breached credential lists against third-party login forms.
• CAPTCHA-solving or bot-detection evasion as a service for fraud. Using hosted sessions to defeat anti-fraud or anti-bot controls on a third-party site in order to commit fraud, create fraudulent accounts at scale, or facilitate ticket/inventory/account-creation scalping in violation of the target site's terms.
• Scraping at abusive scale or in violation of a target site's terms. Reasonable, targeted automation against sites that permit it (or that you're contractually entitled to access) is the product's purpose. Automated scraping that ignores a target's stated terms of service or technical access controls, at a volume intended to burden or degrade the target's service, is not.
• Distributing malware or attacking infrastructure. Using a hosted session as a launch point for malware distribution, exploitation attempts, denial-of-service traffic, port scanning, or any attempt to reach infrastructure you're not authorized to reach. Note that hosted sessions are already technically restricted to public HTTP(S) destinations and cannot reach loopback, private-network, or link-local/metadata addresses (see the Terms of Service §4) — attempting to circumvent that boundary is itself a violation.
• Circumventing platform limits or safety controls. Attempting to bypass storage quotas, instance-hour limits, the vault's origin allowlist or redaction, or a workspace's configured spend-approval policy, rather than requesting a plan change or a policy change through legitimate means.
• Reselling or sublicensing raw access without authorization. Operating AgentBrowser as an unauthorized resold backend for a competing automation service, outside whatever partner/reseller terms (if any) you've separately agreed to.
• Violating others' privacy or intellectual property. Using automation to harvest personal information at scale for purposes that violate applicable privacy law, or to reproduce copyrighted content beyond what the source explicitly permits.
• Testing or probing the Service itself without authorization. Security research against AgentBrowser's own infrastructure is welcome through a responsible-disclosure channel — see the Support Policy — not by attempting exploitation against production without coordinating first.

3. What we do to enforce this technically (not just by policy)

Some of this is enforced by the platform itself, not only by this document:

• Hosted sessions cannot navigate to loopback/private/link-local/metadata network addresses (internal/egress), and customer-configured webhook delivery is subject to the same boundary.
• Vault-injected credentials are redacted from your automation's own output, which also limits accidental credential exfiltration through logs or recordings.
• Storage and instance-hour quotas are enforced against real usage before a write completes, not just reported after the fact.
• If you configure consequence-based spend approvals (P2-207), actions that exceed your policy pause for human approval rather than executing automatically.

None of this is a substitute for your own judgment about what's authorized — it's a partial technical backstop, not a claim that misuse is impossible.

4. Reporting abuse

If you believe AgentBrowser is being used to violate this policy — by another customer, or against your own systems — contact [email protected] with as much detail as you can provide (source, target, timestamps).

5. Consequences

We may suspend or terminate access for a violation of this policy, with notice where practical. Severe violations (attacks on infrastructure, malware distribution, clear fraud) may be suspended immediately without prior notice.