Giving an AI agent its own IP address

Description and use

An AI agent can now exit through a user‑supplied VPN or proxy, eliminating detection from cloud IPs and account conflicts.

Problem

An AI agent that logs into sites on your behalf arrives from a cloud provider’s address range, which sites can detect and treat as non‑human, causing login failures and account merging.

Outcome

Agents now exit from user‑supplied IPs, eliminating detection based on datacenter addresses and preventing account merging caused by shared IPs.

Workflow

• 1. Create or obtain a VPN profile (WireGuard or OpenVPN) from a provider.
• 2. Paste the provider’s config into the workspace’s VPN settings; the config is encrypted and stored only for the workspace.
• 3. For each browser identity, set an override to use a specific VPN profile (or leave unset to inherit the workspace default).
• 4. When a request is made, the system resolves the exit: identity setting → workspace default → direct; if the chosen exit cannot be established, the request fails loudly.
• 5. The VPN tunnel starts on demand, stays idle for ~15 min, and shuts down; the first request after a gap waits ~4 s for the tunnel to come up.
• 6. Browser sessions are rebuilt against the new exit when the override changes, ensuring traffic uses the intended IP.

Evidence

An agent that logs into sites on your behalf arrives from a datacenter, and the sites can tell.

Source

You paste the provider’s config as‑is. It is encrypted with a key scoped to your workspace, and no endpoint returns it, the form is empty when you reopen it because there is genuinely nothing to show, including to you.

Source

If the chosen exit cannot be produced, the tunnel will not come up, the proxy is switched off, the selected profile was deleted, the request fails.

Source