Verifiable provenance, deterministic governance, and zero-trust validation for any agent, on any infrastructure.
Purpose. ATF-AI is a universal, infrastructure-agnostic framework that answers the question:
"How do AI agents prove they are trustworthy, traceable, and auditable regardless of the underlying infrastructure?"
Model. A free and open protocol coordinated by AgroNet Labs. Blockchain, cloud, IoT, enterprise, and any other execution environment can implement ATF-AI as a trust layer independently, without coupling to any specific technology stack.
ATF-AI establishes a governance and trust layer for autonomous agents decoupled from any specific infrastructure. Any system that needs to prove the legitimacy, provenance, and operational integrity of AI-driven actions can implement ATF-AI.
"The wheel already exists.
We're adding autonomous navigation, verifiable provenance, and deterministic governance."
The ATF-AI protocol operates through three infrastructure-agnostic layers:
- Agent Layer Autonomous AI agents performing logic, synthesis, validation, and orchestration tasks.
- Governance Layer Deterministic validation, cryptographic provenance, and zero-trust verification of every agent action.
- Execution Layer Protocol-agnostic infrastructure executing validated workflows across any runtime environment.
Any system "cloud, on-premise, decentralized, or embedded” can implement these three layers using ATF-AI's open specification.
| Pillar | Description |
|---|---|
| Verifiable Provenance | Every agent action is cryptographically signed and traceable via in-toto attestations and OpenTelemetry traces. |
| Deterministic Governance | Validation rules are explicit, reproducible, and auditable no hidden logic, no opaque decisions. |
| Zero-Trust Validation | No agent or system is implicitly trusted. Every interaction is verified before execution. |
ATF-AI is the framework. Specific technology integrations are optional downstream adapters not core dependencies.
| Adapter | Description | Link |
|---|---|---|
| erc-8040-ecosystem | ATF-AI adapter for blockchain/ESG digital asset workflows | github.com/agronetlabs/erc-8040-ecosystem |
| Documentation | Live docs on GitHub Pages | agronetlabs.github.io/atf-ai |
Want to build an ATF-AI adapter for your infrastructure (cloud, IoT, health, fintech, agro)? See CONTRIBUTING.md.
- Open, AI-assisted governance for validation and certification.
- Coordinated through AgroNet Labs, strictly following the Autonomous Trust Framework for Artificial Intelligence (ATF-AI) specification.
- See GOVERNANCE.md for full governance model.
Openly distributed under MIT License.
Implementation and certification trademarks remain under AgroNet Labs governance.
73 tests passing across 4 languages. Zero failures.
| Component | Language | Tests | Status |
|---|---|---|---|
| ERC-8040 Core | Rust | 31/31 | Passing |
| Python SDK | Python 3.12 | 30/30 | Passing |
| C++ SDK | C++17 | 10/10 | Passing |
| Backend (Settlement) | Rust/Axum | 10/10 | Passing |
| Total | 81/81 | Zero failures |
ATF-AI-AUDIT-{SHA256} generated automatically on every settlement operation.
Clean Rust build, 10/10 unit tests passing, server live.
AgroNet Labs LLC
https://agronet.ai
E-mail: admin@agronet.io
Telegram: @agronetlabs