Agent Manifest Concept

Why this exists

Agent Manifest addresses a specific problem:

As AI agents become more autonomous, composable, and networked, much of the surrounding tooling describes how they communicate, what they can do, or how the services they call constrain access. Comparatively little lets an agent’s deployer declare, before interaction and in a neutral, portable form, who it is, what it intends, what it will not do, and who is accountable for it.

Most existing approaches focus on:

Agent Manifest intentionally does none of those.

The Gap

Today, agents can act before any clear declaration of:

This creates opacity by default.

Agent Manifest defines a declarative layer that precedes interaction, rather than following execution.

What This Is

Agent Manifest is:

It is designed to be read by:

Without requiring trust in the runtime.

What This Is Not

Agent Manifest does not:

Those responsibilities belong elsewhere.

The Intent

The intent is not control.

The intent is clarity before action.

Agent Manifest allows any agent to say, in advance:

“This is who I am.
This is what I am for.
These are my limits.”

Nothing more.
Nothing less.

Why Now

The need for declarations precedes:

Agent Manifest is intentionally small so it can:

A Primitive, Not a Platform

Agent Manifest is not a product.

It is a primitive.

A building block that others can adopt, ignore, extend, or reference.

Its value comes from being simple enough to survive.