GetWebStack is a CLI tool and developer platform that automatically analyzes your project and spins up a fully isolated, production-like development environment in a single command. It detects your stack — frameworks, databases, services — and generates all the necessary Dockerfiles, Kubernetes manifests, and routing configuration without any manual setup. Environments run locally or on your own on-premise Kubernetes cluster, meaning your source code never leaves your infrastructure. It scales from a solo developer running one environment to a fleet of AI coding agents each working in parallel sandboxes on the same cluster.

GetWebStack is built for any developer or team that loses time to environment setup and configuration. Solo developers use it to eliminate the daily ritual of starting services manually. Engineering teams use it to ensure every developer works in an environment that mirrors production exactly, eliminating "works on my machine" failures. It is also purpose-built for AI-forward teams running coding agents like Claude, Cursor, or Codex — each agent gets its own isolated sandbox with a real running stack to develop and validate against, rather than coding blindly without feedback.

Setting up a development environment typically takes anywhere from 30 minutes to several days, depending on the project complexity. GetWebStack reduces that to a single command — gws init analyzes your repo and generates everything, then gws up starts your entire stack. Beyond speed, it guarantees production parity: your local environment uses the same infrastructure configuration as your production deployment, so bugs caused by environment drift are eliminated. For teams using AI coding agents, it solves a fundamental problem: agents can now test and verify their own work against real infrastructure instead of making assumptions.

How does zero config work?

When you run gws init in your project directory, GetWebStack scans your repository structure, detects your frameworks, identifies service dependencies like databases and APIs, and automatically generates all the configuration needed to run your environment. This includes Dockerfiles, Kubernetes manifests, environment variables, and domain routing — none of which you write manually. Currently supports all JavaScript, TypeScript, Python, and Go frameworks. Any developer or AI agent cloning the project can run gws up and get an identical environment immediately.

GetWebStack is designed with a local-first, on-premise architecture, which means your source code never leaves your machine or your own infrastructure. There is no cloud dependency and no third-party service that processes or stores your code. When running on a shared on-premise cluster, each developer and AI agent works in a fully isolated Kubernetes namespace, preventing any cross-environment access. This makes GetWebStack suitable for teams with strict data residency requirements or those working on proprietary codebases.

What AI agents work with it?

MCP support is on our roadmap. Currently, AI agents interact with GWS through the CLI directly. gws up -w spins up an isolated sandbox any agent can work against. Join our Discord to follow MCP progress and get early access when it ships.

Can multiple developers share the same cluster?

Yes. When deployed to a shared on-premise Kubernetes cluster, GetWebStack gives each developer or AI agent their own isolated namespace with a full running stack — frontend, backend, database, and routing — all without conflicts. Ten developers can work on ten different features simultaneously on the same cluster, each with their own live URL and independent environment. For local development, each developer runs their own independent cluster instance on their machine. This architecture also supports advanced workflows like PR preview environments, where every pull request automatically gets its own live environment for reviewers to test against.

GetWebstack vs Ona

Two tools built to give developers and AI agents real environments to work against. Same problem space, completely different philosophy.

GetWebStack

Runs on your machine or your cluster
Source code never leaves your infra
One command, zero config
MCP-native, any agent
Flat cost, no metering

Ona

Runs on Ona's cloud or your AWS VPC
Managed infrastructure, remote VMs
Cloud account + devcontainer setup
Proprietary Ona Agent runtime
Consumption-based billing (OCUs)

Architecture

The fundamental difference:
where your code lives

This is the fork in the road. Every other difference flows from this single architectural choice.

GetWebStack

Runs entirely on your infrastructure

GetWebStack spins up isolated Kubernetes namespaces on your local machine or your own cluster. There is no management plane sitting in a third-party cloud. Your source code, secrets, and build artifacts never leave the machine they run on.

Runs on macOS, Linux, Windows locally
Deploy to any existing Kubernetes cluster
No cloud account, no VPC to configure
Works air-gapped by design

Ona

Runs on Ona's cloud or your AWS VPC

Ona provisions remote VMs in its own multi-tenant cloud (US and EU regions) or inside a customer-managed AWS VPC for enterprise accounts. Your environments live on infrastructure Ona orchestrates, not on your local machine.

Multi-tenant cloud (US + EU) or AWS VPC
Remote VS Code in browser or local IDE via extension
Ona's management plane handles auth and coordination
Source code stays in runners, not Ona's control plane

Why this matters for privacy-sensitive teams: If your team works under data residency requirements, handles proprietary codebases, or cannot send source code through third-party infrastructure, GetWebStack's local-first model is the cleaner fit. With Ona, even on a self-hosted VPC, Ona's management plane still handles authentication and coordination.

AI Agent Support

Both support AI agents.
Very differently.

AI coding agent support is where these two tools diverge most sharply in philosophy.

GetWebStack

MCP-native sandboxes, zero lock-in

Install the GetWebstack MCP server once with gws mcp install and any MCP-compatible agent — Claude, Cursor, Codex, Opencode — immediately gets the ability to spin up isolated sandboxes, read logs, and validate its own changes against real running infrastructure.

One MCP server, any compatible agent
Each agent gets its own full isolated stack
No proprietary agent runtime to adopt
Agents work against local or on-prem infrastructure

Ona

Ona Agents — powerful but proprietary

Ona runs its own AI agent runtime inside its environments. Agents write code, run tests, open PRs, and respond to review feedback autonomously. They can be triggered from Linear issues, Slack messages, or Sentry errors. Multiple agents run in parallel across different environments.

Built-in agent runtime (Ona Agent)
Background automations on schedules or triggers
AGENTS.md for project convention guidance
Claude Sonnet 4 as default LLM (enterprise: configurable)

The key tradeoff: Ona's agents are more autonomous and production-ready for background workflows. GetWebstack's approach gives any agent you already use a real environment to work against without adopting a new agent runtime or switching LLMs.

Side by side

Feature comparison

The full picture at a glance.

Feature

GetWebstack

Ona

Infrastructure

Local machine or your own Kubernetes cluster

Ona cloud (US/EU) or AWS VPC

Source code leaves your infra?

✓ Never

~ VPC only (enterprise)

Setup time

One command: gws init && gws up

Cloud account + devcontainer config

Zero config detection

✓ Auto-detects stack

~ Template-based setup

AI agent integration

MCP-native — works with any MCP agent

Built-in Ona Agent runtime

Agent lock-in

✓ No lock-in

✗ Ona Agent runtime

Parallel environments

✓ Per branch/worktree

✓ Per task/environment

Local dev support

✓ First-class

~ Remote IDE primary

PR preview environments

✓ Built-in

Pricing model

Flat — runs on your hardware

Consumption-based (OCUs per task)

Predictable cost

✓ No per-usage billing

~ Variable (OCU usage)

Enterprise SSO / RBAC

~ Cluster-level

✓ Built-in enterprise

Audit logs

~ Infrastructure-level

✓ Full compliance suite

Background agent automations

✗ Not included

✓ Triggers from Linear, Slack, Sentry

Pricing

Very different cost models

GetWebstack runs on your own infrastructure. Ona bills per task completed.

GetWebstack

Pay for the platform, not per environment

GetWebstack runs on your own machine or cluster, so infrastructure cost stays with you rather than accumulating on a third-party bill. There is no per-environment creation fee, no metered compute unit, and no bill-shock risk from an agent running longer than expected.

Ona

Ona Compute Units (OCUs) — per task

Ona bills by OCUs, which cover both environment and agent usage. Creating a new app from scratch costs around 4 OCUs; adding a feature to a mid-size project costs around 8 OCUs. The platform warns that there is "large variability in the number of OCUs consumed per task", which means costs can be unpredictable at scale.

Decision guide

Which one is right for you?

Neither tool is universally better. They serve different teams and different constraints.

Choose GetWebstack if…

✓ Your team cannot send source code through third-party cloud infrastructure
✓ You want each developer and AI agent to have their own full stack, without any infrastructure knowledge required
✓ You want AI agents to work against your local or on-prem stack — using Claude, Cursor, or Codex — without adopting a new agent runtime
✓ You want predictable infrastructure costs with no per-task billing
✓ You're starting from a fresh repo or small team and want the fastest path from clone to running environment
✓ Local development is a first-class workflow, not a secondary concern

Choose Ona if…

✓ Your team is comfortable running environments in a managed cloud or AWS VPC
✓ You want fully autonomous background agents — triggered from Linear, Slack, or Sentry — without managing any Kubernetes
✓ Enterprise compliance is a hard requirement: SSO, RBAC, audit logs, command deny-lists
✓ You want agents to write, test, and ship code entirely autonomously, not just validate changes
✓ Your team works across multiple devices and needs environments accessible from any browser

Try GetWebstack free

No cloud account. No credit card. Your stack, running locally in under 5 minutes.

GetWebstack vs Ona

The fundamental difference:where your code lives

Runs entirely on your infrastructure

Runs on Ona's cloud or your AWS VPC

Both support AI agents.Very differently.

MCP-native sandboxes, zero lock-in

Ona Agents — powerful but proprietary

Feature comparison

Very different cost models

Pay for the platform, not per environment

Ona Compute Units (OCUs) — per task

Which one is right for you?

Choose GetWebstack if…

Choose Ona if…

Try GetWebstack free

The fundamental difference:
where your code lives

Both support AI agents.
Very differently.