Agents are already calling each other. They're just doing it badly.
People are handing errands to AI agents, and businesses are answering with AI agents of their own. Today those two agents usually meet on a phone line or a chat widget designed for humans.
Why Emissar exists
When a person's agent needs a refund, a booking, or a quote, it has three options: call the company's phone number and talk, fill in a web form, or type into a chat window. On the other end there's increasingly an AI agent reading, listening, and replying. Each exchange is converted to speech or prose and back, costs both sides time and money, and leaves neither side able to confirm who it's dealing with.
The protocol for agents to talk directly already exists. A2A, now hosted by the Linux Foundation, defines how two agents exchange tasks. What doesn't exist yet is the network around it: a way to find the right agent from a phone number, to know whether it's trustworthy, to prove what it's allowed to do, and to keep a record both sides accept. That's what Emissar is building.
What we believe
Build on open standards
We don't invent a protocol where one exists. Emissar builds on A2A for agent exchanges, and on AP2 and x402 for payments.
Publish the missing pieces
Where we define something new, such as mandates or signed receipts, we intend to publish the specification openly so anyone can implement it without us.
Say what's real
Every module on this site has a public status, and we change it only when something ships. We won't show customer logos or numbers we don't have.
Keep people in charge
Agents act within limits a person set, and anything outside those limits goes to a person with the full record attached.
Contact
General questions: hello@emissar.ai. More addresses are on the contact page.