Architecture

Control plane in the cloud. Agents in your environment.

HyperLocal's control plane orchestrates intent. AI agents and the ForgeAgent run entirely inside your environment — cloud or on-prem. You choose the LLM. Your code and data never leave your boundary.

HyperLocal Cloud

HyperLocal Control Plane

Orchestrates intent. Does not store or proxy customer data.

SaaS

Governance policies

Cluster configs

Lifecycle orchestration

Compliance intent

Fleet monitoring

orchestrates

Customer-Owned Environment

Your cloud account, your datacenter, your perimeter

Data never leaves

AI Agent Layer

Your choice of LLM

Supported LLM backends — your cloud, your choice

Cloud-hosted LLMs

Azure OpenAI, Vertex AI, Bedrock, or compatible

Self-hosted

Ollama, vLLM, TGI

On-prem GPU

Air-gapped, fully offline

Bring your own

Any OpenAI-compatible API

Assessment Agent

Analyses apps + DBs, recommends architecture

Provision Agent

Infra, K8s, policies, certs via ForgeAgent

Rebuild Agent

Builds open-source replacement apps

Migration Agent

Moves data to purpose-fit stores

Operations Agent

CVE patching, certs, anomaly detection

deploys & operates

ForgeAgent

Linux VM

K8s provisioning

Policy enforcement

Upgrade execution

Telemetry reporting

Your cloud account

Managed K8s

Managed DB

Object storage

Cloud LLM

Second cloud / region

Managed K8s

Managed DB

Object storage

Cloud LLM

On-prem datacenter

K8s

Postgres

MinIO

Self-hosted LLM

Air-gapped / classified

K8s

Postgres

MinIO

On-prem GPU

HyperLocal Control Plane

SaaS-managed governance policies, cluster configs, lifecycle orchestration, monitoring, and compliance intent. Does not store or proxy customer data.

AI Agent Layer (your environment)

Five agents — Assessment, Provision, Rebuild, Data Migration, and Operations — running on your choice of LLM. Cloud-hosted, self-hosted, or on-prem GPU clusters.

ForgeAgent + Data Planes

Kubernetes clusters, app runtimes, and purpose-fit data stores operated in AWS, Azure, GCP, or on-prem datacenters. Data never leaves your boundary.

AI agents in your environment

All AI agents run inside your environment on your choice of LLM — cloud-hosted (Azure OpenAI, Vertex AI, Bedrock, or compatible), self-hosted (Ollama, vLLM), or on-prem GPU clusters for fully air-gapped deployments. Your proprietary code, schemas, and data never cross your network boundary.

Assessment Agent — analyses apps + DBs

Rebuild Agent — builds open-source replacements

Migration Agent — moves data to purpose-fit stores

Operations Agent — patches, certs, anomalies

On-prem agents + data planes

The HyperLocal control plane is SaaS-managed. But everything in your environment — AI agents, ForgeAgent, and data planes — can run fully on-premises, including air-gapped deployments. Self-hosted LLMs on local GPU clusters replace cloud-based AI services. The ForgeAgent operates from cached policy state when connectivity is limited.

SaaS control plane — always managed

On-prem GPU for AI agents

Air-gapped Kubernetes clusters

Local object storage (MinIO)

ForgeAgent

A Linux-based VM deployable via marketplace images or custom enterprise images.

It enables centralized control with distributed regional execution across customer environments. Works alongside AI agents to provision infrastructure, enforce policies, execute upgrades, and report telemetry.

Deployment patterns

Marketplace image in your cloud provider's account
Hardened gold image for regulated / air-gapped environments
On-prem datacenter with self-hosted LLM + GPU cluster
Connected to your existing IAM, network policies, and registries

Agent Decision Transparency

Every agent action is logged, auditable, and visible. You see exactly what happened, when, and why.

OperateAgent detects drift at 2am

Drift detected — config divergence logged with diff

Severity classified (critical / warning / info)

Auto-remediation applied if policy allows

Alert sent to on-call channel with full context

Audit record created with before/after state

Critical drift requires human approval before remediation

RebuildAgent generates a new app

Architecture recommendation logged with reasoning

Each generated file tracked in Git with agent commit

Static analysis (SonarQube) run automatically

Test suite generated and executed

Deploy gate requires human approval

No code deploys without explicit human sign-off

MigrationAgent moves production data

Migration plan generated with row counts and checksums

Dry run executed first — results logged

Differential sync keeps target current

Validation report: row counts, checksums, sample comparison

Rollback available at every step — old env stays live

Production cutover requires explicit human approval

Full audit trail

Every action, every agent, immutable log

Human gates

Critical operations require explicit approval

Rollback-ready

Every state change is reversible

Your observability

Logs, metrics, alerts in your existing tools

See the architecture in action

Request a demo to see how the control plane, AI agents, and ForgeAgent work together in your cloud environment — or fully on-prem.

Request a Demo See Transformations