anewera: the directory where AI agents don't just find businesses — they contact them

If you've made a website agent-ready — navigator.modelContext, llms.txt, structured data — you've done step one: an agent that lands on your domain can act on it. But most agents don't start on your domain. They start with a question: "find me a Swiss relocation agency that can take a corporate brief." Step two is being in the place the agent looks — and being contactable once it gets there.

That's anewera: the network layer that sits on top of individual agent-ready sites.

Visibility is not contactability

A whole category of tools now measures AI visibility — Profound, Peec, Otterly and others show whether and how often a language model names a brand. That's useful, but it's measurement, not action. You end up knowing that you're mentioned, while an agent still can't do anything.

anewera is deliberately not another measurement layer. It's the one after: the place where an agent moves from "I can see this business" to "I've sent this business a request." Different in kind, not degree.

Two kinds of AI, two ways to act

The model is simple and honest:

• Agents that can POST — Custom GPT actions, MCP-connected clients, coding agents like Claude Code — call the API or MCP directly. To contact: prepare_contact_request or a POST to the company's contact endpoint. To list: request_listing.
• Agents that can only read — ChatGPT web, Gemini, Perplexity — build a prefilled link the user confirms with one click.

Both routes share the same backend, spam filter and rate limits. Clean messages are delivered immediately; suspected spam is held. And the business's private recipient email is never exposed — contact runs through anewera. The business decides which actions are allowed.

What makes a profile "agent-ready"

Readers of this blog will recognise the stack — it's the same vocabulary we build with. Per company, anewera ships:

• agent.json and agent.md — the machine-readable description plus a ready-to-use action block.
• an llms.txt entry and a webMCP manifest (/.well-known/webmcp.json), so AI systems find the business and discover its contact functions.
• an MCP endpoint (/api/mcp) with search and contact tools, plus an OpenAPI description for HTTP clients.

If you want the deep dive on that architecture, we wrote it up in Agent-Ready Website Architecture and Agent Actions & API Readiness. anewera is what those techniques point at once you go from one site to a network.

Two markets, one infrastructure

anewera is built dual-domain: anewera.ai serves the English-speaking, global market; anewera.ch serves the German-speaking DACH region. Same application — profiles, knowledge articles and the machine-readable layer appear on both, cleanly separated by language. The knowledge base (anewera.ai/learn) is a good place to start.

Why curated

anewera is early, and it solves a chicken-and-egg problem: infrastructure for the agentic era needs both listed companies and agents that use them. So it currently accepts at most ~30 companies a day — not a marketing trick, but the honest verification limit. In a short call the company data, contact email and allowed actions are confirmed. Two ways in: curated-and-free (bounded by verification capacity) or paid-and-instant — USD 29/month, verified and published immediately, cancel anytime. That's how the directory stays clean: no spam profiles, no invented contact routes, no unconfirmed data.

Where OpenHermit fits

We think about this as two layers of the same shift. OpenHermit is the build layer — the tooling that makes an individual site agent-capable. anewera is the network layer — where agents discover and act across many agent-ready businesses. Make your site agent-ready; then be listed where the agents are looking. anewera.ai is the direct route to becoming contactable.