Background

What A2App stands on.

The orchestration core underneath A2App, the production line it runs, and the case for doing any of it this way. None of it is something you use directly. All of it decides what an agent is allowed to build.

Strategic Infrastructure

Meet AOSF —
the Harness Layer

The Agentic Open Source Foundation provides the orchestration core that makes A2App possible. In the four-layer model of agentic software production, AOSF owns the most strategic position: the Harness — the layer that controls everything beneath it.

01 / SPEC LAYER

Human defines intent Human

Intent, requirements, and acceptance criteria. Where every project begins — the human articulates what needs to be built.

02 / HARNESS LAYER

Orchestration & workflow AOSF Core ★

The strategic control plane. AOSF owns the workflow engine, agent coordination, and process IP. Who controls the Harness, controls the economy. Agents are replaceable — the Harness is not.

Visit AOSF →
03 / AGENT LAYER

AI execution Replaceable

Coding agents, test agents, security agents. The AI workers that carry out tasks — orchestrated by the Harness. GPT, Claude, Gemini: swap freely.

04 / VERIFICATION LAYER

Quality assurance Automated

Static analysis, formal verification, and fuzzing. Every agent output is validated before release — correctness proven, not just tested.

Key insight

We don't own agents — they're replaceable commodities. The model you use today will be superseded tomorrow.

We own the Harness — process IP, governance framework, and quality standards. Strategic, durable competitive advantage.

We invest in token compute — not offices, engineers, or code repos. A true asset-light factory model.

AI Software Factory

Six steps from intent
to release.

The agentic software production pipeline — from a human expressing intent, through AOSF's harness orchestrating agents, to verified and auditable output.

01

Intent

Human describes the goal in plain language

Human
02

Spec

AOSF translates intent into formal requirements & acceptance criteria

AOSF
03

Harness

AOSF Core orchestrates workflow engine and agent coordination

AOSF Core
04

Agent execution

Coding, test, and security agents produce the software

AI Agents
05

Verification

Static analysis, formal verification, and fuzzing — run by AI agents against AOSF quality standards

AI Agents
06

Release

Secure, auditable software ships to production

Output
Sustainable output →Secure, auditable, compliant software — produced without offices, engineers, or code repositories
Why Now

The runtime holds the line.

When AI agents generate interfaces, the runtime does the hard work. Rust's guarantees and Makepad 2.0's declarative model give agents expressive freedom — without giving up the rules.

This is the bet behind A2App and Makepad 2.0: agents describe what the UI should be, and a Rust-native, declarative runtime decides how it behaves. The model gets room to compose. The person using the app gets an interface that stays inside what the runtime allows.