About NessGate

AI systems increasingly need machine-readable entry points into companies — but every company publishes them differently, in a growing pile of standards and locations. An agent that wants to work with a domain has to know about, and check, all of them.

NessGate answers exactly one question:

Given a domain, which machine-readable resources does it
publish — and where?

A resource can be a JSON data file, an API, an agent endpoint (MCP, A2A, or a protocol that does not exist yet), documentation, or a feed. NessGate doesn't host or invent any of this — on each request it reads whatever the domain already publishes across the standard machine-discovery locations (ARD, A2A, llms.txt, API catalogs, OpenAPI, ORD, AWP, host-meta) and returns one normalized answer. Every resource carries a sourceUrl back to the domain, so an agent can go straight to the source and check it directly. The analogy is DNS: DNS resolves a domain to an IP address; NessGate resolves a domain to the machine-readable resources the domain publishes:

DNS       example.com → 93.184.216.34
NessGate  example.com → https://example.com/ai-info.json
                        https://api.example.com
                        https://example.com/mcp

Principles

Resolved on demand, cached at the edge. NessGate does not maintain a copy of your resources. When an agent asks about your domain, NessGate fetches whatever you publish and normalizes it in the response, then caches that answer at the edge for up to 10 minutes. Change what you publish and the next answer reflects it within about 10 minutes — there is nothing to update on our side.

No stored domain data. NessGate keeps no database of domains and no record of what it read. Everything lives on your own infrastructure, under your own name; NessGate holds only a link back to it. It never stores, indexes, or crawls your site, and it never retains the contents it fetched to build an answer.

Neutral — no ranking, no gatekeeping. NessGate does not score, rank, rate, or endorse anyone. It reports what a domain publishes and where; its optional readiness and connection answers are deterministic assessments of that same published metadata — never a ranking, endorsement, or safety claim. It makes no ownership or safety claim about any resource: the domain is the authority, the sourceUrl points straight back to it, and an agent decides for itself.

Open and independently implementable. Resolving a domain is free and public for everyone (AI agents, crawlers, scripts, and humans) through a simple open API with no keys and no sign-up. The behavior is documented in the Spec and the code is open source, so anyone can run their own resolver and get the same answer. NessGate is complementary to the discovery standards it reads, not a competing format.

Bring your own format — and your own protocol. NessGate imposes no required schema, no mandatory fields, and no closed list of resource types. It reads the standards a domain already uses; if a new protocol appears tomorrow, NessGate can resolve its location without needing to understand what is behind it. Agents consume your resources directly from your URLs, so forcing one structure would add friction without adding value.

Deliberately minimal. A resolver should be boring, fast, and predictable. NessGate does one job — turn a domain into one normalized list of the resources it already publishes — and avoids features that would make it a middleman rather than a gate.

Try it

Point it at any domain with the free check, or read the Guide to see how NessGate reads a domain and what it returns. There is nothing to sign up for and nothing to add — if your domain already publishes machine-readable resources, NessGate can already resolve it.