How AI Can Help Platform Engineers Build Better Platforms

© 2026, Amazon Web Services, Inc. or its affiliates. All rights reserved.
How AI Can Help Platform Eng. Build
Better Platforms
Carlos Santana
(he/him)
Sr. Specialist SA, Containers
AWS
https://www.linkedin.com/in/csantanapr
All Things Open RTP Meetup

Packaging / Templating
Code Repository
Config Repository
Artifact Registries
Secret Repository
Signing
Developer Portal
Identity & Access
Secret Management
Validation
Compute Platform
Infra as
code
Continuous
Delivery
Workflow
Orchestration
Service
Discovery
Observability
Terraform
Git
Vault
Argo CD Backstage Argo Workflows Prometheus
External
Secrets
Kubernetes
Internal developer platform capabilities
Kyverno

Cognitive load
Developer experience
Increasing complexity
Operational toil
Challenges of platforms

CNOE is open community collaboration
with the goal to help facilitate platform
engineering through the sharing of guidance,
tooling, and internal developer platform (IDP)
reference architectures
Reduces technology decision
complexity
Leverages community power and
best practices
Provides modular flexibility in
adoption
https://cnoe.io
Cloud Native Operational
Excellence

Evolution into Agentic AI
Generative AI
assistants
Generative AI
agents
Agentic AI
systems
Follow a set of rules
Automate repetitive tasks
Achieve a singular goal
Address broader range of tasks
Automate entire workflows
Fully autonomous
Multi-agent systems
Mimic human logic and reasoning
M O R E
H U M A N
O V E R S I G H T
L E S S
H U M A N
O V E R S I G H T

Developers are
adopting AI
“AI adoption has become nearly
universal. The majority of
survey respondents (90%) use
AI as part of their work and
believe (more than 80%) it has
increased their productivity.” 1
1
“DORA – State of AI-assisted Software Development”, September 2025 (https://dora.dev/research/2025/dora-report/)

27%
Developer time
on innovation and
transformation
73%
Developer time spent
on running and
maintaining applications
< one hour coding time
The median developer spends less than one hour a day writing code1
1 Software.com, “Global Code Time Report”, January 2022

Agent Actions Environment
Observation
Developer portal
IDE / CLI
Workflows
Meet developers where they are
Messaging

Other use cases
Security
remediation
Build with the
platform
Working with platform
abstractions
Triaging and actioning
vulnerabilities
Optimize costs
Provide context-
sensitive
recommendations
Troubleshoot
issues
Faster resolution of
production and CI/CD
incidents

Building AI agents for
platform engineering

Strands
Open source agent frameworks
CrewAI
LangGraph

Software catalog
Cloud infrastructure
Incident management
Cloud costs
CI/CD
Security tools
Documentation
Container orchestration
Context and actions

Model Context Protocol
(MCP)
Introduced by Anthropic, MCP
provides a standardized way
to connect AI models to
different data sources and AI-
powered tools.
Standardized protocol connecting
AI models with external tools
Extensible ecosystem enabling
real-time data and actions
Hosted locally or remotely
Open source protocol

Open source
55+ MCP servers
~ 6.5k stars
1M+ downloads
MCP servers
https://github.com/awslabs/mcp
AWS API MCP
Server
Amazon EKS MCP
Server
AWS Lambda MCP
Server
AWS Step Functions
MCP Server
Amazon DocumentDB
MCP Server
Amazon DynamoDB
MCP Server
AWS Cost Explorer
MCP Server
Amazon Bedrock KB
MCP Server

GitHub
MCP server
GitHub API
Kubernetes API
Backstage
TechDocs
Agents & Model Context Protocol
Kubernetes
MCP server
Backstage
MCP server
AI Agent

Scaling an agent
• More tools can cause agent
confusion
• Longer context can be
truncated
• Unable to optimize for
specific tasks or domains
https://blog.langchain.com/benchmarking-multi-agent-architectures/

Multi-agent patterns
Swarm Supervisor Graph

Agent2Agent
Protocol (A2A)
Enabling AI agents to
discover, communicate, and
collaborate with each other
autonomously to complete
complex, multi-step tasks.
Autonomous discovery and
communication
Collaborative task completion
Standardized protocols
Open source protocol

Agent Framework
LLM
Agent
MCP
Agent Framework
LLM
Agent
MCP
A2A
{
"name": "My Agent",
"description": "I can do cool stuff",
"url": "https://agent.example.com",
”version": "1.1.2",
"skills":[{
"id": "something",
"name": "Something",
"description": "Does something cool",
"examples": [
"Please do something cool for me",
]
}],
"securitySchemes":{
"oidc": {
"openIdConnectUrl": "something",
"description": "Internal auth"
}
}
}
A2A in practice

Multi-agent AI
Platform
Agent
CI/CD Agent
AWS API
Catalog Agent
Golden Paths Agent
Argo CD
GitHub
Backstage
IDE / CLI
Messaging

Multi-agent AI
Platform
Agent
CI/CD Agent
AWS API
Catalog Agent
Golden Paths Agent
Argo CD
GitHub
Backstage
IDE / CLI
Messaging
IDE / CLI
MCP
Platform
Agent

Multi-agent AI
Platform
Agent
CI/CD Agent
AWS API
Catalog Agent
Golden Paths Agent
Argo CD
GitHub
Backstage
IDE / CLI
Messaging
IDE / CLI
A2A
CI/CD Agent
Catalog Agent
Golden Paths Agent
Platform
Agent

Multi-agent AI
Platform
Agent
CI/CD Agent
AWS API
Catalog Agent
Golden Paths Agent
Argo CD
GitHub
Backstage
IDE / CLI
Messaging
IDE / CLI
A2A
CI/CD Agent
Platform
Agent

Multi-agent AI
Platform
Agent
CI/CD Agent
AWS API
Catalog Agent
Golden Paths Agent
Argo CD
GitHub
Backstage
IDE / CLI
Messaging
IDE / CLI
MCP
Platform
Agent
AWS API
CI/CD Agent

Multi-agent AI
Platform
Agent
CI/CD Agent
AWS API
Catalog Agent
Golden Paths Agent
Argo CD
GitHub
Backstage
IDE / CLI
Messaging
IDE / CLI
MCP
Platform
Agent
Argo CD
CI/CD Agent

Multi-agent AI
Platform
Agent
CI/CD Agent
AWS API
Catalog Agent
Golden Paths Agent
Argo CD
GitHub
Backstage
IDE / CLI
Messaging
IDE / CLI
MCP
Platform
Agent
GitHub
CI/CD Agent

Building agents with open source
Protocols Agent frameworks
Enabling standardized
communication between
agents, allowing them to
share context and co-
ordinate tasks.
Accelerate development
with pre-built components
for orchestration, state
management and inter-
agent collaboration.
MCP servers
Integrate with popular data
sources and tools to quickly
connect agents to existing
systems without custom
integrations.

Case studies:

Outshift by Cisco is the incubation
engine delivering what’s next and
new for Cisco.
Agentic AI Quantum

AWS – CI/CD
Outshift platform
Edge
Workloads
Control
Security
Active security Vulnerability scanning
Atlantis
AWS
GPUs
QPUs
Secret management
and
more
Argo CD Backstage

• Interact with JIRA
• Bootstrap new source repository with CI
• Setup CD for application
• Provision EC2 developer instances
• Enable access to LLMs
• RAG: Platform Docs
• Live tool calling: Kubernetes, Argo CD
• Graph RAG: AWS, Backstage
Complexity
Agentic Platform
Engineering
Outshift Developer
JARVIS
Webex
Backstage
JIRA
CLI
VS Code
K N O W L E D G E B A S E S
S E L F - S E R V I C E
• Kubernetes deployments to Outshift sandbox
with prompts
C O D E G E N E R A T I O N

Open Source Projects

Open source, multi-
agent system for
platform engineering
Secure, scalable, persona-driven reference
implementation
Built-in knowledge base retrieval to streamline
platform operations
Integrations for popular platform engineering tools
like Kubernetes, Argo CD, Backstage and more.
Drive agent interactions from CLI, developer portal
Built on open source technology including A2A,
MCP, Agntcy, LangGraph and OpenTelemetry
https://caipe.io

Frameworks:
LangGraph, Strands Agents
Protocols:
MCP, A2A
Argo CD
Backstage
GitHub
AWS
PagerDuty
Confluence
Jira
Incident Engineer
Product Owner
CAIPE
Open source
Agents Personas
Platform Engineer
CAIPE
Slack
Orchestrators:
Docker Compose, Kubernetes
Splunk Webex
Observability:
OpenTelemetry

Starting with
knowledge
• Lowers barrier to entry
• Higher quality outcomes
• More capable AI systems

CAIPE: Unified RAG architecture
• Build a knowledge graph as a
RAG source for agents
• Ingests data from data
sources like AWS, Kubernetes
& Backstage
• Ontology agent automatically
discovers relationships

BYO
Framework

Kagent
• CNCF Sandbox Project
• Open Standards (A2A, MCP)
• Kubernetes Native (CRDs)
• Observability
https://kagent.dev

MCP
LLM
Agent
Agent
Agents
Skills
System Prompt

mcp-target
agentgateway
MCPServer
MCPServer
secretsRefs
server ontainer
Agentgateway integration (agentgateway.dev)

mcp-target
agentgateway
RemoteMCPServer
RemoteMCPServer
HeadersFrom

How AI Can Help Platform Engineers Build Better Platforms

More Related Content

Similar to How AI Can Help Platform Engineers Build Better Platforms

More from All Things Open

Recently uploaded

How AI Can Help Platform Engineers Build Better Platforms