AWS Builder Center
AINE-DEV - The Enterprise Agentic AI Toolkit on AWS

AINE-DEV - The Enterprise Agentic AI Toolkit on AWS

AINE.DEV is the Technology Toolkit within the broader AINE (AI-Native Enterprise) framework — AWS's Enterprise Agentic AI Transformation program, that enables engineers to build enterprise grade Agentic applications.

What is AINE.DEV?

AINE.DEV is the Technology Toolkit within the broader AINE (AI-Native Enterprise) framework — AWS's Enterprise Agentic AI Transformation program. While AINE addresses the full spectrum of enterprise AI adoption spanning strategy, people, data, risk, and operations, AINE.DEV is the engineering layer that turns AI ambition into production reality. AINE-DEV delivers AWS Cloud-Native Agentic AI Blueprints, a pre-built agent library on composable 10-layer architecture to guide enterprises adopting AI natively across their processes. Agnostic of harnesses used by the enterprises, proven patterns are converted into reusable accelerators that can be pre-loaded into customer’s harness of choice so that engineering teams configure rather than construct.
AINE.DEV is an Intelligent Harness — an MCP-native, LLM-agnostic framework that packages AWS AI programs as portable skills and steering, selects the right programs per customer based on their intention (industry, maturity, goals), applies industry-specific regulatory lenses, and renders governance-aware artifacts natively into whatever agentic platform the customer already uses.
"The durable asset isn't the agent — it's the harness. Agents are interchangeable; the AWS programs that steer them are the value."

Three Core Promises

#
Promise
What It Delivers
Key Metrics
1
Freedom to Build
Most comprehensive agentic AI toolkit with model choice, multi-framework support, and LLM-agnostic architecture
42 Accelerators | 117 Agentic Samples13 Platform Adapters | LLM-Agnostic
2
AI You Can Trust
Governance-first architecture with pre-wired security pillars embedded across all application layers
7 Governance Pillars Across 10 LayersAudit Trails | Compliance-Ready
3
Maximize Value
3-5x faster delivery with proven PoC-to-production conversion across 23 industries
3-5x Faster Delivery80% PoC-to-Production | 23 Industries

How It Works

AINE.DEV operates as an MCP (Model Context Protocol) server — a single hosted endpoint that any agent platform connects to.
The workflow is simple and repeatable:
Step 1 — Describe: Load an intention.yaml capturing the customer's industry, AI maturity level, and business goals.
Step 2 — Resolve: The harness evaluates activation predicates and composes the right programs with industry lens overlays.
Step 3 — Render: Native artifacts flow into the agent platform of choice — Kiro, Claude, Cursor, Copilot, Amazon Q, and 8 more.
Step 4 — Build: The agent executes the steering and skills; the harness governs the output — auditably.

One Sentence to a Harness

An intention is a plain-language description of what your customer needs. Paste your customer's sentence to your agent, it loads it via fde_load_intention, resolves it, and renders the harness. The kit evaluates your intention against every program's activation rules and turns on exactly the ones that match, then overlays your industry lens.


The kit operates through loop engineering: programs don't just activate — they feed each other during execution. Architecture decisions inform requirements, requirements inform code generation, RAI reviews produce guardrail configurations that get implemented, and resilience patterns get deployed. Each iteration enriches the system. The agent doesn't just follow a plan — it refines the plan as it builds.
Think of the kit as an accelerator for your FDE pod, enabling human-AI collaboration across the full delivery lifecycle. You describe a customer's situation (industry, goals, AI maturity), and the kit activates the right programs, applies industry-specific lenses, and orchestrates an interleaved delivery — from requirements capture through working code and deployment.
This article covers the architecture, show how programs self-activate based on your project context, and demonstrate how to go from a YAML intention file to a fully scaffolded AI application on AWS — complete with governance gates, responsible AI reviews, and multi-platform rendering. The following figures illustrate the key architectural concepts:
An engineer defines their engagement in a declarative intention.yaml file, or as a free flow text, which feeds into the Core Engine (Loader, Planner, Renderer, State Manager). The engine activates relevant programs from a catalogue of 9 and renders platform-native artifacts for Kiro, Claude Code on Amazon Bedrock, and Codex on Amazon Bedrock.

The AINE-DEV Catalog

The following table summarizes all 9 programs in the catalogue:

Program
Purpose
Best Practices
AIM
AI MaturityAssessment — six-perspective scoring, routes downstream
AWS Gen AI Maturity Model
AI-DLC
AI-Driven Development Lifecycle. Three-phasemethodology (Inception → Construction → Operations)
awslabs/aidlc-workflows
AI Native App Builder
Scaffolds AI-nativeapps using the 9-layer architecture with CDK and Strands agents
26 reference patterns
AgentPath
Agentic AI strategy, architecturepatterns, human-in-the-loop design
AWS Agentic AI Patterns
Responsible AI
Five-dimension RAIreviews, guardrails, model cards, red-team exercises
AWS Responsible AI
AI Operations
Enterprise operating model, AI BOM,MLOps pipelines, incident response
AWS Prescriptive Guidance
AI Resilience
Fault-tolerance,blast-radius mapping, chaos engineering
AWS Well-Architected Reliability
AI-PLC
Product development lifecycle with KPItaxonomy and measurement
aws-samples/sample-ai-plc
FDE Orchestration
Meta-program thatcoordinates all others
FDE Delivery Kit

The AI Native AppBuilder Architecture

The apps you build with the AINE-DEV Toolkit follow the AI Native AppBuilder architecture: a layered architecture for AI-native applications on AWS. Each layer is independently deployable, with skills and accelerators mapped to every tier.
Cross-cutting concerns such as Governance & Security, and Observability run across all layers.


Industry Lenses: Compositional Governance

The lens system is what makes AINE.DEV an intelligent harness rather than a static scaffolding. When the customer's industry matches a lens, regulatory and domain-specific governance automatically overlays onto active programs — no manual rewriting required.
Example: A generic Responsible AI program delivers five-dimension reviews, model cards, and red-team methodology — solid, but industry-blind. Compose it with the financial-services lens and the output transforms: SR 11-7 model risk governance, SOX/ICFR controls, PCI-DSS, regulatory reporting, and separation of duties. A banking-grade program — from a single industry field.
Industry Lens
Regulatory & Domain Overlays
Financial Services
SR 11-7 model risk governance, SOX/ICFR controls, PCI-DSS, regulatory reporting, separation of duties
Healthcare
HIPAA, PHI protections, SaMD classification, clinical safety
Life Sciences
GxP, clinical trial data governance, FDA/EMA compliance
Manufacturing
Industrial safety, OT/IT convergence, supply chain integrity
Public Sector
Digital sovereignty, data residency, classification
Games Industry
Player data protection, real-time performance, content moderation

What is unique about the AINE.DEV Harness

Most agent toolkits are pipelines: configure once, execute linearly. The FDE Kit runs delivery loops. Most agent toolkits are pipelines: configure once, execute linearly. The AINE-DEV Kit runs delivery loops where each program's output enriches the system and feeds the next:
Load intention → Resolve programs → Build
↑ ↓
Re-resolve ← Patch intention ← Discover
↓
Implement findings → Ship
Three mechanisms make this work:
1. Shared state mutation — Programs call  aine_patch_intention  as they discover new aspects  from the customer intention. When  ai-plc  working-backwards produces "build a claims processing agent", it patches the intention's  workload.primaryDescription . Downstream programs see enriched context without being told.
2. Artifact handoffs — Every skill has explicit handoff routing. RAI (Responsible AI) review findings become guardrail specs → and the build implements them as Bedrock Guardrails. Blast radius maps become degradation policies → the build implements circuit breakers. Output from one program is input to the next program progressively.
3. Re-resolve cycles — After  sync-to-aine-intake  or state transitions, the system calls  aine_resolve  again. New programs can activate mid-engagement based on what was discovered (e.g., discovering  regulated: true  activates  responsible-ai ).
The result: the agent doesn't just follow a plan — it refines the plan as it builds. Each iteration produces working code, not documents about what to build.where each program's output enriches the system and feeds the next:

One Kit, Multiple Engagement Types

The same kit supports multiple engagement shapes:
Engagement Type
Programs Activated
Typical Duration
Discovery Workshop
AIM only
1 day
PoC Sprint
AI-DLC + AI Native App Builder
1–2 weeks
MVP Build
AI-DLC + AI-PLC + RAI+ AI Native App Builder + AI Ops
3–4 weeks
Production Deployment
All programs + Industry Lens
6–8 weeks
Operational Review
AI Operations + AI Resilience
2–3 weeks
You configure the engagement type in the intention file, and the resolver activates the relevant programs — helping avoid over-engineering for a workshop or under-governing for a production set-up

Proven Outcomes

+80%PoC-to-Production Conversion
3-5xFaster Delivery Timeline
~2.5 wksvs. 12+ Weeks Traditional
23Industries Proven
Customer wins have been across industries.
  • Financial Services: Invoice processing reduced from 3 days to 4 hours. VPC-isolated, retained as production system.
  • Manufacturing: Stockout events reduced by 40%. 12-month backtest, board-level business case.
  • Telecommunications: 40% Tier-1 query deflection. Live 30-day trial, expanded to Tier-2.
  • Philippines Bank: 98% time reduction, 50x velocity increase in document processing.
  • Global Credit Ratings: 8 months of work delivered in 1 week
  • Leading Media Network: 6 apps in 2 days, 4x productivity, 223 engineer-days saved.

Positioning Summary

Positioning Summary
AINE-DEV
What
Technology Toolkit for AI-Native Enterprise Transformation
Part of
AINE — AWS's Enterprise Agentic AI Transformation Framework
Architecture
Intelligent Harness — MCP-native, LLM-agnostic, agent-interchangeable
Key Differentiator
Industry lenses auto-compose regulatory governance without manual rewriting
Delivery Methodology
AI-DLC: Spec-driven, 3 weeks to production
Platform Support
13 agentic platforms (Kiro, Claude, Cursor, Copilot, Amazon Q, +8 more)
Governance
7 pillars across 10 layers — auditable, deterministic output
Target Audience
Chief AI Officers, Application Developers, Security Experts, GRC Teams
Open Source Reference
github.com/aws-samples/sample-fdep-mcp-server (MIT-0 Licensed)

AINE-DEV is part of the broader AINE (AI Native) Framework

AINE.DEV is the technology toolkit within the broader AINE program — AWS's Enterprise Agentic AI Transformation Framework. AINE addresses seven critical challenges in AI adoption: business case and value creation, data strategy and governance, security and compliance, measurement and evaluation, change management, skills gaps, and adoption.
AINE.Capabilities
AINE.DEV
AINE-ENGINEER (example FDE)
Diagnostic Engine 47 capabilities, 6 domains Surfaces gaps, maps to programs
Technology Fulfillment using 42 accelerators, 117 samples Converts gaps into deployables
Delivery Layer. Forward-deployed engineersRapid prototypes in days



Reference Implementation of AINE-DEV

The AINE-DEV Toolkit gives you a reference implementation for Agentic AI with multi-platform rendering out of the box. Fork the repository, extend it with your own delivery patterns, and brand it as your own. Everything you build on top remains your intellectual property. The reusable patterns, lenses, and delivery assets you create compound over time — each engagement makes the next one faster.
The AINE-DEV toolkit is made available as the aine-mcp server,  which helps forward-deployed engineers build AI-native applications on AWS from tested architecture patterns. Describe a customer's situation (industry, goals, AI maturity), and the kit activates the right programs, applies industry-specific lenses, and guides the agent through a structured delivery — from requirements through working code and deployment.
The kit doesn't just configure an agent workspace — it drives actual application development using 26 verified reference patterns (CDK stacks, Python agents, Bedrock configuration, React UI) and the AWS AI-DLC methodology (Inception → Construction → Operations). Programs interleave during delivery: architecture decisions feed requirements, requirements feed code generation, RAI reviews produce guardrail configs that get implemented, resilience patterns get deployed.
Reference Implementation of AINE-DEV
The Partner FDE Delivery kit is a reference implementation of the AINE-DEV Toolkit. The FDE delivery toolkit is being used by partners to build and deploy customer Agentic AI applications. The fde-mcp-server source code is opensourced with MIT0 license and is available for partners and customers to extend, modify and implement their own Agentic solutiions. [AWS Sample] https://github.com/aws-samples/sample-fdep-mcp-server.git   

Start Your AINE.DEV Journey

Customers can start their AINE.DEV journey by contacting their Account team. The engagements for AINE can vary based on the customer context, and can range from a half-a-day workshop to upto 3weeks of engagement.
Step
Engagement
Duration
Outcomes
1
AI Maturity Assessment (AIM)
Half Day
47-capability gap analysis across 6 domains
2
AI RoadmapWorkshop
1-2 Days
10-layer framework mapping, acceleration points identified
3
Rapid Prototyping Engagement by partner FDE or )CDE
5-10 Days
Working agent deployed in customer environment
4
Production Roadmap
Weeks 2-3
Spec-driven path from prototype to governed production

Conclusion

The AINE-DEV kit aims to bridge the gap between AI experimentation and enterprise production. By combining structured methodology (AI-DLC), self-activating programs, industry compliance lenses, and multi-platform rendering, it delivers working applications through loop engineering — programs that feed each other, state that evolves as the agent builds, and tested AWS patterns that ensure production-grade output. The result is a repeatable, governed path from intention to deployed AI workload.
Whether you're building a claims-processing agent for a regulated bank, a RAG assistant for a healthcare provider, or a multi-agent system for an entertainment company — the AINE-DEV Tookit supports methodology, governance, and platform rendering, freeing your team to focus more on delivering customer outcomes.
Contact your AWS Account Team to begin  |  github.com/aws-samples/sample-fdep-mcp-server  |  MIT-0 Licensed

Authors

 Shivagami Gugan  , Chief Technologist at AWS, the author of the AINE-DEV framework specializes in building and implementing Agentic AI transformations and large-scale modernization programs for governments and large enterprises. She is also the Vice Chair of the Steering Committee at the United Nations AI for Good initiative
 Byron Arnao   is a Principal Technologist at AWS and Global Responsible AI Lead, is the co-author of the AINE-DEV framework, he specialiizes in architecting the strategic and technical frameworks that define how enterprise AI is built, evaluated, and governed. He focuses on advancing human-AI co-intelligence.
 Phani Lingamallu   is a Principal Partner Solutions Architect at AWS. He is the author of the fde-mcp-server, the partner toolkit of the AINE-DEV framework. He works with strategic partners to design and deliver AI-native solutions using structured methodology, agentic frameworks, and the FDE Delivery Kit.

Contributors

 Dhana Vadivelan   is a Senior Manager at AWS leading EMEA Partner Solutions Architecture for strategic consulting (GSI) partners. He drives the FDE and AI-DLC partner programs, accelerating AI workload delivery through structured methodology, agentic development platforms, and joint go-to-market strategy with AWS consulting partners.
Any opinions in this article are those of the individual author and may not reflect the opinions of AWS.
Enjoyed reading this content? Let the author know!

Your likes, comments, shares, and saves help creators reach more builders.

Loading recommendations

Loading article