The Enterprise AI Engineering Framework for building production-ready software with AI coding agents.
naeos
The Enterprise AI Engineering Framework for building production-ready software with AI coding agents.
README
NAEOS
Nusantara AI Engineering Operating System
The Engineering Control Plane for AI Coding Agents.
Build software with AI agents under explicit architecture, policy, execution, evidence, and verification controls.
NAEOS is an open-source engineering framework for building production software with AI coding agents. It provides the engineering layer between what a team intends, what an agent proposes, what is authorized to execute, and what can be independently verified afterward.
It is not another AI coding assistant. NAEOS is the control plane around the agent.
Why NAEOS?
AI coding agents can generate code quickly. Production engineering still requires explicit answers to harder questions:
- What architecture is the agent operating within?
- Which actions are allowed?
- Who or what authorizes an external side effect?
- What actually happened during execution?
- What evidence proves the result?
- Can another system independently verify the claim?
- Can the same engineering intent be carried across different AI agents?
NAEOS treats these as engineering-system concerns rather than prompt-writing concerns.
The model
Without an engineering control plane:
Prompt → AI Agent → Code / Action → ?
With NAEOS:
Specification
↓
NEIR
↓
Validation + Policy
↓
Agent Context / Intent
↓
Authorized Execution
↓
Observation → Evidence
↓
Independent Verification
Two architectural views describe the same system:
Engineering model
Specification → NEIR → Validation → Policy → AI Context → Generation
AI Engineering control plane
Agent Intent → Policy Decision → Authorized Execution
→ Observation → Evidence → Independent Verification
The first describes how engineering knowledge moves through NAEOS. The second describes how agent actions are controlled and made auditable.
Core concepts
| Concept | Role |
|---|---|
| Specification | Declares system intent and engineering requirements |
| NEIR | Canonical engineering representation derived from specifications |
| Policy | Determines which proposed actions are permitted |
| Runtime | Executes only within the authorized boundary |
| Evidence | Records observable results and integrity-relevant metadata |
| Verification | Independently checks claims about outputs or side effects |
| Handoffs | Carries explicit contracts between agents, tools, and execution boundaries |
| AI Compiler | Translates engineering context into agent-specific instructions |
| Extensions | Adds profiles, plugins, adapters, and integrations without changing the core model |
See it run
The fastest way to understand NAEOS is the CLI control-plane demo:
git clone https://github.com/NAEOS-foundation/naeos.git
cd naeos
go build -o naeos ./cmd/naeos
./examples/demo-cli/run-demo.sh
The demo exercises the specification-first pipeline and produces traceable run metadata.
Golden Path: docs/GOLDEN-PATH.md defines the reproducible five-minute flow, acceptance criteria, and evidence map used for technical evaluation and regression testing.
Reference Demo & Evidence Story: docs/REFERENCE-DEMO.md turns the same run into an independent reviewer checklist, traceability narrative, and partner-pilot evidence format.
External Validation: docs/EXTERNAL-VALIDATION.md provides a reproducible third-party runbook and evaluation record. Engineering Note #001: docs/ENGINEERING-NOTE-001-EXTERNAL-VALIDATION.md explains the engineering rationale and evidence boundary for independent external validation.
External Validator Program: docs/EXTERNAL-VALIDATOR-BRIEF.md is the public entry point for Cohort #001; see Program and Missions for the program protocol.
For the Todo API demonstration:
go build -o naeos ./cmd/naeos
./examples/todo-api/run-demo.sh
See START-HERE.md for the supported onboarding path.
A concrete run
The naeos run path exposes structured engineering-run metadata, including identifiers such as:
run_id
specification_hash
neir_hash
validation_state
policy_evaluation
generated_context
artifacts
evidence
The goal is not merely to generate code. The goal is to preserve enough engineering context and evidence to understand what was intended, what was authorized, what happened, and what can be verified.
AI coding agents
NAEOS is designed to remain vendor-neutral at the engineering layer.
NAEOS targets Go 1.26.10 or later and can produce agent-specific instruction/context artifacts for tools such as:
- GitHub Copilot
- Claude Code
- OpenAI Codex
- Cursor
- Gemini CLI
- OpenCode
- Windsurf
The engineering model remains upstream of the individual agent. Changing agents should not require rebuilding the project's engineering rules from scratch.
Evidence and verification
NAEOS includes engineering experiments and implementation paths around:
- policy boundaries and authorization
- durable audit/evidence records
- tamper detection
- handoff contracts
- independent verification
- artifact signing and verification
- SBOM generation
- security and vulnerability checks
- benchmark and fuzz gates
These experiments are evidence of specific mechanisms and behaviors. They should not be interpreted as blanket proof of every production property.
See experiments/ for the current experiment suite, including the Evidence V5.6 milestone.
Trust model
NAEOS is built around explicit boundaries:
- Intent is not authorization.
- Policy decisions are distinct from runtime execution.
- Observations are distinct from claims.
- Evidence should outlive an agent's conversational memory.
- Verification should not depend solely on the component making the claim.
- Untrusted handoff input must not silently become downstream authority.
- Controllable boundaries must be revalidated before consequential external actions.
- Version and contract mismatches should fail closed where required by the governing specification.
These are engineering principles, not guarantees that every integration is automatically safe.
Architecture authority
NAEOS uses a normative documentation hierarchy:
Constitution
↓
Reference Architecture
↓
Master Technical Specification
↓
Engineering Specifications / ADRs
↓
Implementation
↓
Experiments
Public guides such as README, START-HERE, and GETTING-STARTED explain and navigate the system; they do not override normative specifications.
Start with DOCUMENTATION-AUTHORITY.md, then review:
- NAEOS Reference Architecture
- NAEOS Master Technical Specification
- NAEOS Specification
- NEIR
- CLI Reference
Architecture Drives Engineering.
Repository map
naeos/
├── cmd/naeos/ # CLI
├── internal/
│ ├── specification/ # parsing, normalization, resolution
│ ├── neir/ # engineering representation
│ ├── compiler/ # AI instruction compilation
│ ├── governance/ # policy and governance
│ ├── runtime/ # execution gateway
│ ├── evidence/ # evidence storage
│ ├── verification/ # independent verification
│ ├── audit/ # audit trail
│ ├── signing/ # artifact signing
│ ├── pluginsdk/ # plugin SDK
│ └── ... # supporting subsystems
├── pkg/ # public Go packages
├── docs/ # engineering specifications
├── experiments/ # executable engineering experiments
├── constitution/ # project constitution
├── governance/ # governance material
└── examples/ # runnable examples
Current release
NAEOS 3.7.1 is the current documented software release.
NAEOS software releases and experiment milestones use different version sequences. For example, NAEOS 3.7.1 is a software release, while Evidence V5.6 identifies an engineering experiment milestone.
For release history and changes, see CHANGELOG.md.
Documentation
- START-HERE.md — onboarding and first contribution path
- GETTING-STARTED.md — developer setup
- DOCUMENTATION-AUTHORITY.md — normative documentation model
- WHITEPAPER-EN.md — English whitepaper
- WHITEPAPER.md — official English whitepaper
- DOCUMENTATION-INDEX.md — document index
- docs/ — NAEOS engineering specifications
- CHANGELOG.md — release history
Contributing
NAEOS is built in the open. Contributions are welcome across the engineering stack:
| Track | Examples |
|---|---|
| Policy | authorization, governance, policy evaluation |
| Runtime | execution boundaries, lifecycle, runtime controls |
| Evidence | audit, receipts, integrity, evidence storage |
| Handoffs | agent/tool contracts and boundary revalidation |
| Verification | independent checks and trust mechanisms |
| Plugins | SDK, adapters, profiles, extensions |
| Experiments | reproducible engineering experiments and benchmarks |
| Documentation | specifications, architecture, guides |
Read CONTRIBUTING.md, then start with START-HERE.md.
Slack + GitHub community
NAEOS uses Slack for real-time engineering coordination and GitHub as the durable engineering source of truth.
- NAEOS Slack Community — real-time engineering discussion
- GitHub repository — source code, issues, pull requests, and evidence
For workspace administrators, install the official GitHub integration for Slack, then subscribe the relevant channels to repository activity. The integration supports GitHub notifications, issue and pull-request collaboration, rich link previews, and GitHub actions from Slack. See GitHub's Slack integration documentation.
Recommended operating model:
Slack coordinates → GitHub records → CI verifies → Evidence closes the loop
For public community acquisition, use the workspace's Slack shared-invite URL rather than the workspace home URL. Keep that invite URL configurable in the website deployment so it can be rotated without changing repository code.
Community and partnerships
- GitHub Discussions — technical discussion and ideas
- Discord — community discussion
- NAEOS Partner Program — technology, infrastructure, AI, academic, community, pilot, and strategic collaboration
Social identity and distribution
Social Identity Source of Truth defines the canonical public positioning, channel roles, profile copy, visual rules, and verification status for NAEOS social distribution.
Security and governance
- SECURITY.md — supported release lines and security reporting
- Engineering Constitution — project-level engineering principles
- Governance — project governance material
- AI-assisted provenance policy — guidance for AI-assisted contributions
- Open-core boundary — relationship between NAEOS Core and future offerings
- Legal Architecture — IP provenance, trademarks, third-party licensing, and release legal controls
License
NAEOS is released under the Apache License 2.0. See NOTICE for attribution and branding information.
Contributions are accepted under the Developer Certificate of Origin as described in CONTRIBUTING.md.
The NAEOS, NEIR, and NAEOS logo names are branding of the NAEOS OSS. The Apache License grants no trademark rights; see NOTICE for the current branding and trademark policy.
Active development · NAEOS 3.7.1
Architecture Drives Engineering.