Zendesk Launches Public Directory for AI Agent-to-Agent Service Discovery
Zendesk has launched a public directory that lets AI agents discover which companies support agent-to-agent (A2A) service protocols, aiming to solve a key discoverability gap in automated customer service.
Zendesk has launched a public directory designed to help AI agents locate agent-to-agent (A2A) service options when resolving customer requests. The directory acts as a discoverability layer, allowing autonomous AI agents acting on behalf of customers or businesses to identify which companies support A2A protocols and route service interactions accordingly.
The move positions Zendesk at the center of an emerging interoperability layer for AI-driven customer service, where one company's AI agent can hand off or coordinate tasks with another company's service systems without a human initiating each step.
As AI agents increasingly act on behalf of customers — booking, canceling, troubleshooting, or escalating issues autonomously — those agents need a reliable way to discover which businesses can actually be reached and serviced through agent-to-agent channels. A public directory solves a basic but previously unaddressed problem: discoverability. Without it, AI agents are left guessing which companies support machine-to-machine service interactions, limiting how far automation can scale across the customer journey.
For transformation leaders, this signals a shift from AI agents operating inside a single brand's walls to AI agents operating across brands, platforms and vendors. Service design now has to account for a new "customer" that is itself a software agent, with its own expectations around structured, discoverable, machine-readable service options.
This launch is less about a single feature and more about infrastructure — the plumbing that determines whether AI-to-AI service actually works at scale, or stalls as a collection of disconnected pilots.
Most organizations are still designing AI agents to talk to humans, not to other agents — and that's the blind spot this directory exposes. The real behavioral shift isn't customers trusting AI; it's companies having to make their service options legible to machines they don't control. Any operator serious about agentic CX should audit today whether their service endpoints are even discoverable in a standard, machine-readable way — because if an AI agent can't find you, it will route the customer elsewhere, and no one will ever know a request was lost.
