From answering to acting.
ANOTS is honest about what is live versus what is coming. Here is the line between the two — and where the architecture is headed.
Shipped, in production, answering today.
Running on the GYIBB pilot right now — every item here is something an AI can hit live.
Public agent endpoint
Every tenant gets a discoverable A2A agent card at anots.com/<slug>/.well-known/agent-card.json, and an MCP endpoint alongside it.
Live answers from your own data
A 90-second chat crawls your site into a knowledge base. The agent answers questions from it — and from product tools — instead of letting AIs guess from your HTML.
The go-live tag
A copy-paste package (llms.txt section, .well-known bridge, HTML head block) you drop on your own domain so visiting AIs discover your agent. The "GA tag" for the AI internet.
The ANOTS directory MCP
Install one endpoint and reach every business on ANOTS. search_businesses, ask_business, and find_and_ask let an AI discover the right agent and ask it — in one step.
Next out the door.
Active work. The plumbing exists; we are wiring the control surface and the first outbound paths.
Alert — outbound signals
When a high-value question, escalation, or anomaly comes in, ping Slack, Google Chat, email, or any webhook. Phase 1: webhook + email.
Report — operator digests
A daily digest in your channel of choice: top asks, conversions, cost, what the AIs wanted that you did not have.
Control tower permissions
Pick inbound tools (catalog, FAQ, pricing) and outbound ones from one place, with an audit log on every call.
Where it goes.
The vision the architecture is built toward. Directional, not dated.
Act — agents that do, not just say
Book a meeting on your calendar, file a ticket, send a follow-up, initiate a refund. Bind the agent to live systems — CRM, stock, scheduling — so the answer is "I have it, here is the price, the slot is Thursday."
Custom domains
Serve your agent from agent.yourbusiness.com instead of a path on ours. Your brand, your URL.
A directory that ranks for truth
When find_and_ask fans out to many agents, a trust layer adjudicates between them — grounded in the synthesis and multi-agent arbitration work (TCAM) underneath ANOTS, so the aggregated answer is honest, not loudest.
Peer-to-peer A2A
Agents that hand off to each other directly — your agent calling a supplier’s agent calling a logistics agent — a real network of business agents, not a hub of isolated ones.
The honest summary: today your agent is found and answers. The roadmap is about teaching it to act — and connecting one agent to all the others. See pricing or explore the directory MCP.