What is WebAZ An Agent-Native Protocol Experiment for the AI Era

#ai #agents #opensource #mcp

AI makes one person more capable than ever.

But capability is only half of the story.

Commerce access, contribution records, reputation, evidence, and accountability are still mostly locked inside platforms. If a person uses agents to do real work, where does that work become visible? Who can verify it? Who remains accountable when the agent acts?

That is the question behind WebAZ.

WebAZ is an early agent-native commerce and contribution protocol for the AI era.

It is not another marketplace pitch. It is not a token story. It is not a reward promise. It is a protocol experiment around one simple design direction:

Humans should be able to use a normal app.
Agents should be able to use a structured interface.
Both should connect to the same underlying rules.

In WebAZ, the human-facing surface is the PWA. The agent-facing surface is MCP.

That split matters.

A human can browse, buy, sell, contribute, and review state through the PWA. An agent can search, read public state, propose actions, or execute bounded workflows through MCP. Both paths point to shared concepts: products, orders, tasks, contribution records, evidence, and accountability.

The goal is not to make agents "in charge." The goal is to make agent-assisted participation legible.

Why this exists

Most platforms were designed for humans clicking buttons.

AI agents change the shape of work. A single person can now ask an agent to inspect a repo, draft a fix, compare products, summarize requirements, or prepare a contribution. That creates new power, but also new ambiguity.

If an agent places an order, what rules did it follow?
If an agent contributes code, how is that contribution attributed?
If a stranger starts with one small task, how can the system remember that participation without pretending it is a guaranteed reward?

WebAZ starts from a conservative answer:

Record participation.
Attach evidence.
Keep boundaries explicit.
Avoid reward promises.

This is why contribution in WebAZ is treated as attribution and accountability first. A contribution record is not a payout claim. It is a way to make real work easier to see, verify, and remember.

The current mental model

The simplest way to understand WebAZ is:

Humans use the PWA.
Agents use MCP.
Shared rules connect both sides.

The PWA is for people. It gives users a familiar interface for commerce and participation.

MCP is for agents. It gives agents a structured way to interact with the same system without scraping screens or guessing workflows.

The shared rule layer is the important part. It keeps the system from becoming two separate worlds: one for humans and one for machines. The same order state, task state, evidence trail, and accountability model should remain visible from both sides.

This is also why WebAZ is careful with language. "Agent-native" does not mean "agent-ruled." Agents can execute bounded work, but humans or organizations remain accountable.

What WebAZ is not

WebAZ is pre-launch.

It is not a mature production economy.
It is not an investment product.
It is not a token promise.
It is not a guaranteed reward system.

Those boundaries are deliberate.

The current promise is narrower: make real participation easier to see, verify, and remember in an AI-assisted environment.

Why MCP matters here

MCP gives agents a more structured interface than browser automation or ad hoc scraping.

That is important for commerce and contribution workflows because agents need boundaries:

What can be read?
What can be proposed?
What can be changed?
What requires human confirmation?
What evidence should be attached?
What state transition happened?

For an agent-native protocol, the interface is not only a convenience. It is part of the trust boundary.

WebAZ uses MCP as the agent-facing doorway, while the PWA remains the human-facing doorway.

A small example

Imagine an open-source contributor starts with one small task:

one doc fix
one test
one bug report
one small PR

That work might begin outside WebAZ, such as on GitHub. Later, the contributor may want to claim identity or connect that contribution to their broader participation record.

The important thing is not to turn every contribution into a payout promise.

The important thing is to avoid losing the work.

WebAZ is exploring how contribution can be recorded first and claimed later, while keeping evidence and boundaries visible.

The direction

If WebAZ works, the end state is not a giant platform that owns everyone.

The end state is a thinner protocol layer where humans, agents, sellers, buyers, and contributors can coordinate through shared rules.

People should be able to participate without disappearing into a platform database.
Agents should be able to help without becoming unaccountable actors.
Small contributions should be remembered without being oversold.

That is the experiment.

Code is Rule.
Protocol is Trust.

If you are interested in the idea, start with the WebAZ site, the short explainer video, or the public project materials.