About Agentonomy

Mission

Agentonomy is for people who build, secure or bet on artificial intelligence (AI) agents. It gives them an honest, evidence-graded picture of the infrastructure agents need to act in the economy. That means identity, permissions, payments, protocols, verification, security, operations, provenance and governance.

Every page answers one question: what works in production, what is only a specification, what is announced, and what is still unsolved?

What Agentonomy is

Agentonomy is the independent maturity map of AI-agent infrastructure: every claim dated, every rating evidenced. Its signature asset is the Ladder, a scale built to rate named, versioned items from L0 (concept) to L5 (standardised and boring). The rules behind every rating are set out in the methodology, published at launch together with the first ratings.

Newsletters report news, directories list servers, and foundations promote their own projects my analysis. We have found no other publication that rates maturity across every layer of the agent stack with a public evidence trail my analysis. That gap is the reason this site exists.

It is written first for the Builder, the engineer wiring agents into real systems, and the Gatekeeper, the security, identity or platform architect. Technical leads, founders and analysts use it as a one-glance map.

The nine pillars

Protocols and interoperability: how agents talk to tools, services and each other.
Identity and authentication: who this agent is, and how that can be proved.
Delegation and permissions: what an agent may do, on whose behalf, and how that is revoked.
Payments and commerce: how agents pay, buy and settle.
Verification, evaluation and reputation: whether it worked, and whether to trust it next time.
Security and containment: how to stop agents being hijacked or doing damage.
Operations and reliability: how to run agents in production.
Data, content and provenance: what agents read, how it is licensed, and where a claim came from.
Governance and liability: who is accountable when an agent acts, and what the law requires.

What we will never publish

Daily news recaps or “this week in AI” round-ups.
“Top 10 AI agents” lists, tool recommendations or productivity hacks.
Business-idea, side-hustle or “make money with agents” content.
Token, stablecoin or crypto price talk; investment advice.
Model benchmark horse-races.
Predictions without a stated basis, or market sizes without a primary source.
Sponsored content, affiliate links or vendor-written posts.
Tutorials that only repeat vendor documentation. Our hands-on labs exist to test claims.

Editorial principles

Status before story. Every page opens with what is real today, labelled, before any explanation.
Primary sources or it does not run. Specifications, standards drafts, official documentation, release notes and regulators. A press article is a lead, not a source.
Everything is dated. Each page shows when it was last verified; each status label names the version or date it refers to.
Specification, implementation and adoption are different things. We keep them apart in every sentence. Announced partners are not adoption.
Say what it does not prove. Every mechanism we explain comes with a line on what it does not guarantee.
Vendor neutrality. Vendor statements are labelled as vendor claims, and any relationship with a vendor is disclosed on the page.
Corrections in public. Errors are fixed within 48 hours of confirmation, with a dated note on the page and an entry in a public corrections log.
Living pages, not dead posts. State-of pages are updated in place with a changelog. Analysis pieces are dated, frozen and later superseded by a link.

How AI is used here

AI agents monitor sources, score incoming changes, summarise specifications and draft text. A separate checker agent, never the one that drafted, tests each claim against a stored copy of its source.

The editor picks the topics, opens every cited source, checks every factual claim, assigns every Ladder rating and approves every page. Ratings are never assigned by AI alone. Every published article carries this line: “Researched and drafted with AI assistance; every claim verified and approved by the editor on [date].”

Independence

Agentonomy takes no sponsors, uses no affiliate links and publishes no vendor-written posts. There is no monetisation in the first year. No one can pay for a place on the Ladder or for a review, now or later. If any funding model is ever introduced, it will be described on this page before it starts.

Who runs it

Agentonomy is run by one person, the editor, with AI assistance. The editor is an analyst with a background in data and chemistry. The site is method-first: the rules come before the ratings.