Speakeasy and Tailscale Secure Enterprise AI via Private Networking

Many businesses currently rely on public endpoints for AI agent communication, creating potential security vulnerabilities. This partnership allows companies to host the Speakeasy MCP gateway as a node within their existing Tailscale network. Because the gateway is only reachable from trusted devices and identities, administrators can close public routes entirely, ensuring that AI agents operate under the same security constraints as human employees.

Sagar Batchu, CEO of Speakeasy, argues that agent actions require the same oversight as human staff. The integration leverages the Tailscale Kubernetes operator to provision proxies that verify each request before it reaches an MCP server. For firms like payment infrastructure provider Polar, this setup bridges a significant gap in auditing and governance. Polar now keeps its internal support and operations servers private, relying on the tailnet to mediate access for its engineering teams.

This architecture also complements Speakeasy’s existing MCP tunnels, which facilitate outbound-only connections for servers hidden behind firewalls. By combining these, organizations can maintain complex AI workflows without exposing sensitive internal tools to the open web. The feature is available immediately for enterprise users, marking Tailscale as the first networking provider supported by the platform.