Lyzr for AWS

AgentCore governance gets better with Lyzr OpenController.

Keep building on AWS. Add one layer above every cloud you run.

Compare all 317Book a session

Every capability is one cell

  • Yes
  • Partial
  • Roadmap
  • Not offered

Lyzr

303of 317 · +8 partial

AWS

73of 317 · +141 partial

Of 317 capabilities, Lyzr has 303 generally available and 8 partial. AWS has 73 generally available, 141 partial and 103 not offered.

capabilities, scored row by row
317
Lyzr products, each against its closest AWS services
7
capabilities with no AWS equivalent
103
AWS status as of
18 Oct 2026

Seven products. Lyzr leads in all seven.

Each bar is one product’s capabilities, against the closest AWS services. Select a product to see its rows.

Where AgentCore governance stops

OpenController against AgentCore Policy, Gateway, Identity, Evaluations, Agent Registry and Bedrock Guardrails.

135 / 141
generally available in OpenController
58
with no AWS equivalent
  • Lyzr, generally available
  • AWS, generally available
  • AWS, counting partial
Share of each OpenController area that is generally available, Lyzr against AWS. Table view in the comparison below.

0%none of the area is available

100%every capability is available

0%50%100%
  1. Developer devices (Edge)
  2. Deploy
  3. Discover
  4. Posture and evidence
  5. Skills
  6. LLM Gateway
  7. Platform
  8. Tools and MCP
  9. Observability and cost
  10. Register
  11. Evals
  12. Control
  13. Identity
  14. Guardrails
  15. Refine

Right column: generally available capabilities, Lyzr vs AWS. Biggest gap first.

All features compared side by side

Every capability, Lyzr against AWS. You choose what to compare: a product, a bucket, or a single feature.

  1. 1Choose a product
  2. 2Click a colour to filter
  3. 3Search a feature
  4. 4Fold areas, or go full view

OpenController· 141 capabilities. How they compare:

Try

CapabilityLyzrAWS

OpenController · vs AgentCore Policy, Gateway, Identity, Evaluations, Agent Registry, Bedrock Guardrails

  • Databricks serving endpoints, Agent Bricks, Genie, MCP services, Unity Catalog skillsLyzr: YesAWS: Not offered
  • Anthropic and Claude Platform: models used, API keys, workspacesLyzr: YesAWS: Not offered
  • Salesforce, SAP and HubSpot agent platformsLyzr: YesAWS: Not offered
  • Claude Enterprise and ChatGPT Enterprise plansLyzr: YesAWS: Not offered
  • API keys found per provider and linked to the models they callLyzr: YesAWS: Not offered
  • A reason required to reject or revokeLyzr: YesAWS: Not offered
  • Entra agent identities governed next to AWS identitiesLyzr: YesAWS: Not offered
  • Grader agreement against human labels (Cohen’s κ)Lyzr: YesAWS: Not offered
  • Judge token budgets that are never overspentLyzr: YesAWS: Not offered
  • Quarantine and teardown of a versionLyzr: YesAWS: Not offered
  • Block across Copilot Studio, Agent 365, Foundry and SnowflakeLyzr: YesAWS: Not offered
  • Tool groups that keys and teams can useLyzr: YesAWS: Not offered
  • Find agents calling MCP servers outside the registryLyzr: YesAWS: Not offered
  • Find unauthenticated tool gateways, static-key tools and destructive toolsLyzr: YesAWS: Not offered
  • Tool-call checks inside Claude Code before the tool runsLyzr: YesAWS: Not offered
  • One spend view across AWS, Azure, GCP, Databricks and Anthropic billsLyzr: YesAWS: Not offered
  • Licences and idle seats in dollarsLyzr: YesAWS: Not offered
  • Shadow spend: what unregistered agents costLyzr: YesAWS: Not offered
  • ‘Lethal trifecta’ check: outside input, private data and a way to actLyzr: YesAWS: Not offered
  • Agents acting with their maker’s credentials; keys made by leaversLyzr: YesAWS: Not offered
  • Agents calling models around the gatewayLyzr: YesAWS: Not offered
  • Agents nobody traces; traced agents nobody evaluatesLyzr: YesAWS: Not offered
  • One policy for Claude Code, Codex, Copilot and GeminiLyzr: YesAWS: Not offered
  • Scoped virtual keys on laptops, never raw provider keysLyzr: YesAWS: Not offered
  • Claude Code governed through one managed-settings file, nothing installedLyzr: YesAWS: Not offered
  • Govern AWS, Azure, GCP, Databricks, Snowflake and Anthropic from one deploymentLyzr: YesAWS: Not offered

Agent Studio · vs Bedrock, AgentCore, Bedrock Knowledge Bases, Amazon Connect Customer

  • Validators for OpenAPI schemas, SQL and gibberish textLyzr: YesAWS: Not offered
  • Video avatars from HeyGen, Simli and TavusLyzr: YesAWS: Not offered
  • Isolated environments: test clones of an agentLyzr: YesAWS: Not offered
  • One agent deployed to AgentCore and Google Vertex AI Agent Engine togetherLyzr: YesAWS: Not offered

SuperFlow · vs Bedrock Flows and AWS Step Functions

  • Decide node: many typed questions about an item answered in one small-model callLyzr: YesAWS: Not offered
  • Rank node: reorders a list by relevance in one callLyzr: YesAWS: Not offered
  • A completed test run converts into a webhook, schedule or file triggerLyzr: YesAWS: Not offered

Alpha / Architect · vs Apps in Amazon Quick, Kiro (successor to Amazon Q Developer)

  • Voice agents built from plain languageLyzr: YesAWS: Not offered
  • Agents defined as git repositories (GitAgent), portable to other runtimesLyzr: YesAWS: Not offered
  • Marketplace to publish, browse and preview appsLyzr: YesAWS: Not offered

Open CoWork · vs Amazon Quick (successor to Amazon Q Business)

OpenFineTuner · vs Bedrock model customization and SageMaker AI

  • Deduplication at two gates: documents first, then training pairsLyzr: YesAWS: Not offered
  • Routing policy published as a versioned, readable document that goes through approvalLyzr: YesAWS: Not offered
  • Every routing decision logged with the candidates, evaluation evidence and priceLyzr: YesAWS: Not offered

Optimus · vs AWS AI Factories and AWS Outposts

  • Inference with no per-token charge on owned hardwareLyzr: YesAWS: Not offered
  • Same Lyzr software as in the cloud: Agent Studio, Architect and OpenControllerLyzr: YesAWS: Not offered
How it’s scored, and the AWS sources · 28 documents

AWS status as of 18 October 2026. Generally available counts as yes. Scope-limited, preview, beta or sample code counts as partial. The same rule applies to both columns. Re-check a status before you quote it, especially after AWS re:Invent (30 November to 4 December 2026).

Agent CI/CD ships today. On AWS, each team builds its own.

AWS gives you versions, endpoints and A/B tests. Its own guidance has you assemble the pipeline from CodePipeline, CodeBuild or GitHub Actions.

Release stages from development to production. Lyzr: 7 of 7 built in. AWS:1 built in, 4 partial,2 not offered.
  1. StageLyzrAWS

  2. Version

    Model, prompt, tools per version

    Immutable versions

  3. Environments

    Dev → stage → prod

    Endpoints, no promotion record

  4. Eval gate

    Golden-set floors, regression

    A GitHub Actions sample

  5. Approval

    Author can’t self-approve

    Generic manual approval

  6. Canary

    Built in

    Traffic split, no auto rollback

  7. Rollback

    One step

    Repoint an endpoint

  8. Any cloud

    Foundry, Vertex, Copilot Studio

    AWS only

In productionLyzr runs agent CI/CD for PepsiCo.

AWS has not announced a native agent pipeline.

AWS changed direction on three agent products in 2026.

  1. 30 Jul 2026

    Bedrock Agents (classic)

    AgentCore

    Maintenance mode. Closed to new customers.

  2. 31 Jul 2026

    Amazon Q Business

    Amazon Quick

    Closed to new customers.

  3. 30 Apr 2027

    Amazon Q Developer IDE plugins

    Kiro

    End of support. New sign-ups closed since May 2026.

A layer above the cloud keeps your agents, policies, versions and audit history in place when a provider changes its products.

One layer above every cloud you run

AWS keeps the compute. Adding a second cloud later means adding a connection. No agent has to move.

People use one Lyzr layer, deployed in your account, which builds, governs, routes and improves agents on AWS, Microsoft Azure, Google Cloud, data and SaaS platforms, and on-prem.
  • Employees
  • Business teams
  • Developers
  • IT, risk and finance

Every request

Lyzr · one layer, deployed in your account

Work

Open CoWork

AI workspace for every employee, on any model

Build

Alpha / Architect

Agent Studio

SuperFlow

Builders, agent platform and workflows

Govern

OpenController

Discover, register, release, control, observe, prove

Models

LLM Gateway

OpenFineTuner

Routing, budgets, fine-tuned open models

  • AWS

    • AgentCore
    • Bedrock
    • Agent Registry
    • EKS and ECS

  • Microsoft Azure

    • Foundry
    • Copilot Studio
    • Agent 365

  • Google Cloud

    • Vertex AI Agent Engine
    • Gemini Enterprise
    • Agent Registry

  • Data and SaaS

    • Databricks
    • Snowflake
    • Salesforce · SAP

  • On-prem

    • Lyzr Optimus appliance
    • Your data centre
    • Air-gapped sites

Agents keep running where they run. Nothing migrates.

  • What AWS does well

    Strong building blocks for agents that live on AWS

  • Where it stops

    Controls apply inside AWS, one account and region at a time

  • What Lyzr adds

    One layer for every cloud and every department

Runs inside your AWS account, on AWS’s own services

Every action Lyzr takes shows up in your own AWS logs.

Your AWS account

Lyzr

  • Agent Studio
  • OpenController
  • LLM Gateway

  • Deploys onfrom Agent Studio

    • CloudFormation
    • VPC
    • ECS Fargate
    • DocumentDB
    • Aurora
    • Redis
    • CloudWatch

  • Runs onfrom OpenController

    • EKS
    • Secrets Manager

  • Routes tofrom LLM Gateway

    • Bedrock models

    Adds budgets, routing and fallback

  • Deploys agents tofrom Agent Studio

    • AgentCore Runtime

    The same agent can also go to Vertex AI Agent Engine

  • Reads and blocksfrom OpenController

    • AgentCore
    • Agent Registry
    • Bedrock

    Through a cross-account role you control. A block adds a tagged statement to the AgentCore resource policy

  • Shows up infrom OpenController

    • CloudTrail

    The operator of every action appears there

  • Usesfrom Agent Studio

    • AgentCore Memory
    • Bedrock Guardrails

    Next to Lyzr’s own checks

  • Connectsfrom Agent Studio

    • S3
    • Redshift
    • Amazon Neptune

    Knowledge, data sources and graph stores

  • Joins spend fromfrom OpenController

    • Cost Explorer

    With agent traces, licences and other clouds

Stays on AWS

  • AgentCore managed runtime
  • AgentCore Memory
  • Code Interpreter and Browser
  • Identity token vault
  • Bedrock models
  • GuardDuty AI Protection
  • Compute, storage and networking

Lyzr adds, across every cloud

  • Discovery across clouds
  • Governance
  • Release gates
  • Model routing
  • Spend control
  • Builders
  • Workflows
  • Employee workspace
  • Fine-tuning
  • On-prem hardware
  • Buy through AWS Marketplace private offers
  • AWS Generative AI Competency Partner

Bring the AWS agent you worry about most

Book a working session

sales@lyzr.com

  1. 1

    Connect

    A read-only role in your AWS account

  2. 2

    Find

    The agent you worry about most

  3. 3

    Register

    Owner, purpose and approval on record

  4. 4

    Govern

    Release, control and stop it, next to your agents on other clouds

One working session. The agents stay where they run.