Uploaded January 2026 | Updated September 2026, 3 hours ago

---

Remy explains how Tailscale’s private, identity-aware network enables a new way to proxy AI requests without distributing API keys, while giving companies deep visibility and fine-grained controls over both users and agents. They also dive into advanced use cases like running Claude Code inside your Tailnet, isolating workloads with ephemeral Tailnets, and using TSNet to let applications themselves join a private network.

Topics include:

• What makes Tailscale’s AI Gateway different from other LLM gateways
• How identity travels with every request across your Tailnet
• Rate-limiting, quotas, and model access control based on users and groups
• Running agents safely without leaking API keys
• The power of TSNet and programmatic access to Tailnets
• Ephemeral Tailnets for isolated workloads, customer-specific networks, and AI environments
• Why private networking is becoming essential infrastructure for AI systems