✦ Artifact-based test automation

Artifact-based
test automation.
Powered by AI.

test-tracker → test-designer → test-debugger → test-executor. A pipeline of specialized AI agents inside VSCode. Runs in your cloud. Your LLM keys stay with you. You ship runnable Robot Framework — not black-box scripts.

Book a Demo → See the Pipeline
5 platforms GA — Win · Swing · Oracle Forms · Web · Linux
4 AI agents · one pipeline
0 infrastructure to manage
AutoMagic AI Test Automation
Works with
🤖 Claude Code 🐙 GitHub Copilot 📋 TestRail 🎯 Jira / Zephyr 🔧 Jenkins 🐙 Git 🖥️ Windows & Linux ☕ Java Apps 🤖 Claude Code 🐙 GitHub Copilot 📋 TestRail 🎯 Jira / Zephyr 🔧 Jenkins 🐙 Git 🖥️ Windows & Linux ☕ Java Apps
The Agent Pipeline

Four agents. One pipeline.
Artifacts all the way down.

Four specialized AI agents, one VSCode-native environment. You steer — they design, run, debug, and file bugs. Every step produces a reviewable artifact in your workspace.

test-tracker
picks what to automate
📋ticket / spec
test-designer
writes the test
📄.robot file
test-debugger
self-heals locators
🖼️annotated screenshot + fix diff
test-executor
runs at scale · files bugs
📊report + visual log + ticket
🎯
test-tracker

Knows what's covered, what's flaky, what's next. Orchestrates the pipeline by picking the next scenario to automate — from your TestRail, Jira, Zephyr, or suite state.

→ Coverage & state view

🪄
test-designer

Turns a TestRail ticket, Jira story, or plain-English spec into a Robot Framework test. Structured, readable, runnable.

→ Runnable .robot file

🔧
test-debugger

Runs the test, inspects failures, self-heals locators, proposes diffs. Pauses for a human when it gets genuinely stuck — no silent hand-waving.

→ Annotated run log + fix diff

🚀
test-executor

Runs the suite at scale. Produces logs, screenshots, rewindable visual logs, and files bugs in your tracker on regressions.

→ Report + screenshots + visual log + ticket

Why artifacts — not pure autonomy?

Every real AI agent system keeps a human in the loop. The ones that claim otherwise just hide the supervision inside prompt chains nobody reads. The only thing worth arguing about is whether that loop is efficient. Artifact-based means you correct an agent by marking up a .robot file or a screenshot — not by reading pages of reasoning. Narrow, predictable control points. The iterative code → debug → refine cycle that test engineering already runs on stays intact.

Every output is a real file in your workspace — executable Robot Framework code, screenshots, visual logs you can play back, reports, terminal logs, filed tickets. No black-box SaaS. No proprietary format. No vendor lock-in.

What It Automates

Every mainstream enterprise surface.
Including the ones browser-only tools can't reach.

One driver framework — AnyUIDriver — covers desktop, Java, Oracle Forms, web, and Linux from a single client.

Platform Windows Web Java Swing Oracle Forms Linux GUI JavaFX macOS Java macOS Native
Status GA GA GA GA GA Beta GA Beta
Platform Status
Windows desktop apps GA
Web (Chrome / Edge) GA
Java Swing GA
Oracle Forms GA
Linux GUI GA
JavaFX Beta
macOS — Java / Swing apps GA
macOS — OS-level (non-Java) apps Beta

Don't see your platform? Talk to us — we add drivers when there's a real enterprise need.

Architecture

Like BrowserStack —
but for your whole app.

BrowserStack gives you browsers in the cloud. We give you full Windows and Linux machines — run any desktop app, Java Swing, Oracle Forms, or web app in your own private, isolated cloud environment. One tenant. Three layers: AutoMagic on top, LocAItor in the middle, AnyUIDriver at the bottom.

🏠
One environment per customer — fully isolated

Each subscription gets its own dedicated, isolated cloud environment. Not shared infrastructure. Your test data, your apps, your code — isolated from everyone else.

⚡
Zero infrastructure on your side

No QA lab hardware to buy, configure, or maintain. No IT tickets. Workstations, tooling, and test runners all live in the cloud — you just connect and work.

🔗
Your app's backends stay reachable

If your desktop app needs to connect to internal servers, we configure a secure site-to-site or client VPN tunnel between your network and your cloud lab. Your front-end in the cloud sees your back-end on-prem.

🌐
Internet via your machine — no firewall headaches

Your cloud workstations route internet traffic through your local machine. No need for your IT team to open or manage firewall rules on the cloud side — your policies apply automatically.

🤖
Sandboxed enclosure for AI agents

AutoMagic runs as a sandboxed Docker Dev Container: connect from VSCode, pull your repo in, work, push tests back to Git, disconnect. Agents live inside the sandbox — they cannot reach your host machine or corporate network unless you explicitly allow it. If an agent misbehaves, the blast radius is the container. That containment is what makes it safe to give AI agents real control over a UI at all.

Managed by us, controlled by you. Your environment is provisioned automatically on subscription. You get admin access — domain-join machines, add users, customize. We handle the infrastructure; you handle your work.

Your Infrastructure
💻
Your Laptop
VSCode + Plugin
🗄️
App Backends
DB, API, Services
🏢 Enterprise Friendly
🔒 Secure Network Connection (VPN when needed)
⇅
Your Private Cloud Lab
🖥️
Windows Workstation
Your app deployed here · VSCode Dev Container
🐧
Linux Workstation
Web / headless testing · CI runners
Three-layer stack — pre-deployed
🪄
AutoMagic
Agent pipeline + Inspector
🎯
LocAItor
Element search engine
🖱️
AnyUIDriver
Per-platform REST drivers
📊
Reports · Visual Logs
Reviewable artifacts
Alliedium Inspector

Inspector — for the whole desktop.
Not just the browser.

Every web developer knows their browser's element inspector. We built the same thing for native desktop apps — Windows, Java Swing, Oracle Forms, Linux GUIs — right inside VSCode. Hover, highlight, drill into the element tree, grab a stable locator in one click.

🎯
Hover → highlight → tree

Exactly how it works in Chrome. Except it works on native Windows UI, Java Swing dialogs, Oracle Forms screens, and Linux apps too.

📋
One-click locator copy — SmartID

XPath · component metadata · OCR · image. Our multi-strategy identifier we call SmartID — stays stable when the UI shifts, backed by a 95% locator-stability track record over 2 years.

👁️
Shared view with the AI

The agents use the same Inspector to decide what to click. You use it to verify their choices and debug runs. No translation layer between human and agent.

🧩
Ships inside AutoMagic

A first-class pane inside the AutoMagic VSCode plugin — not a separate tool to install, learn, or license.

Alliedium Inspector
For Every Role

One tool. Every role on the QA team.

Same product, same artifacts. Engineers get deeper reach, testers get automation without code, managers get a fully managed lab — everyone contributes to the same test suite.

🧪
Automation Engineer

Reach apps your current tools miss

Already have automation? AnyUIDriver plugs in to cover Windows desktop, Java, Oracle Forms, and web — all in one. LocAItor keeps your locators resilient so you stop chasing broken XPaths.

  • Works alongside existing Selenium or similar setups
  • Robot Framework output by default — or Java, C#, and others on request
  • CI/CD pipeline templates included
📝
QA Team — Limited Coding

From test cases to automation — no code required

Write test cases in plain English (or keep them in TestRail). AutoMagic turns them into well-structured automated tests. You review. You approve. Done. The underlying structure is clean and readable if you ever want to dig in.

  • Natural language → working automation
  • Review and approve before committing
  • Tests run automatically on your cloud machines
📈
QA Manager / Team Lead

Your QA lab — without the lab

No hardware budget. No setup overhead. Your entire test environment — workstations, automation, and reporting — lives in the cloud. Spin up instantly, scale as needed, pay only for active workspaces.

  • No QA lab hardware to procure or maintain
  • Full visibility: video, logs, defect filing
  • One predictable monthly cost per workspace
Security & Trust

Your data. Your keys.
Your network.

We host the platform. You host everything that matters. No shared compute, no AI vendor lock-in, no data leaving your tenant.

📋
Reuse your existing LLM license

GitHub Copilot, Claude Code, or any compatible LLM — we integrate with what you already pay for. No new AI vendor to evaluate or approve.

🏠
Your data stays in your environment

Every customer gets an isolated, private cloud environment. Your test data, your app screenshots, your code — stays in your dedicated space, not shared infrastructure.

📦
Tests you can take with you

Your generated tests are standard Robot Framework. The only dependency is our drivers. If you ever leave, you take almost all your automation code with you.

Supported LLM Providers
🐙
GitHub Copilot
Primary integration · enterprise-sanctioned
🧠
Anthropic Claude Code
Fully supported · same agent pipeline
Most enterprise teams already run Copilot. If you do, adopting Alliedium adds zero new vendor contracts, zero new legal reviews — it's the path of least resistance through your procurement process.
LLM-agnostic by design. The pipeline routes to whichever provider you configure via the AI_BACKEND flag.
Pricing

Enterprise pricing —
tailored to your deployment.

Per-tenant, per-tier, with onboarding and support included. Contact us for a quote.

Alliedium Enterprise
Dedicated Tenant Subscription
One dedicated VPC · everything included · onboarded by our team
Talk to Sales →
Sizing and tier pricing: available on request
Default Workspace (Standard)
● 4 vCPU / 8 GB RAM
● Windows Server workstation
● 100 GB SSD storage
Need more compute? Larger machine options available — contact us.
Every deployment includes
  • ✓ Dedicated VPC per customer
  • ✓ The full agent pipeline (test-tracker, test-designer, test-debugger, test-executor)
  • ✓ LocAItor element-search engine
  • ✓ AnyUIDriver — Windows, Swing, Oracle Forms, Web, Linux
  • ✓ Alliedium Inspector + VSCode plugin + Dev Container
  • ✓ TestRail / Jira / Zephyr integrations
  • ✓ CI/CD pipeline templates (Jenkins, GitLab CI)
  • ✓ Per-run video playback + screenshots + visual logs
  • ✓ Automatic defect filing
  • ✓ Network connectivity setup (VPN as needed)
  • ✓ Onboarding & setup by our team
💡 Bring Your Own LLM

Already running GitHub Copilot or Claude Code? We plug into what you've got. No new AI contracts, no new vendor approvals. Your key stays in your Dev Container.

🏢 Prefer on-premises — or an outright purchase?

All Alliedium components can run in your own infrastructure, fully air-gapped if needed — same tooling, same agents, your hardware. And if a subscription model doesn't fit your procurement, we're open to a perpetual license as an exception. Talk to us.

Talk to Us →
🔬 Not sure yet?

Book a live demo. We'll run the pipeline against one of your real test cases — so you see exactly what the agents produce for your app.

Book a Demo →
See It In Action

Watch AutoMagic work.

30 seconds. Real test. Real output.

▶
Demo video coming soon — book a live demo instead →
Book a Live Demo See the Pipeline
FAQ

Common questions.

No. On subscription, we provision your dedicated tenant automatically — Windows and Linux workstations, the agent pipeline, LocAItor, AnyUIDriver, Postgres, Traefik — all pre-deployed. You get admin access. No cloud account, no IT ticket. Workstations route internet through your local machine by default, so your organization's existing firewall policies apply without any changes on the cloud side.
Yes. We configure a secure network connection (site-to-site VPN or client VPN) between your corporate network and your tenant. Your workstations can reach internal DBs, APIs, and application servers as if they were on your LAN.
GitHub Copilot is the primary integration — most enterprise customers already run it as their sanctioned AI chat platform, so no new vendor approval or contract is needed to adopt Alliedium. Claude Code is fully supported via the same agent pipeline. The pipeline itself is LLM-agnostic through the AI_BACKEND flag. You bring the license and the key — we don't resell AI, we integrate with the subscription your team already pays for.
Windows desktop apps, web (Chrome / Edge), Java Swing, Oracle Forms, and Linux GUI are all GA. On macOS: Java / Swing apps run on the cross-platform Swing javaagent (GA), and non-Java desktop apps are supported in beta via the image + OCR visual driver. JavaFX is in beta. See the platform matrix for the driver technology behind each.
Your tests are plain .robot files in your Git repository. Your locators live in a PostgreSQL database you can export. You can run the entire suite with standard Robot Framework after we're gone. No proprietary formats, no vendor lock-in — that's the whole point of "artifact-based."
Directly. The VSCode extension syncs the workspace to your local machine; commit and push from there. Or download changes as a git patch via the command palette (Alliedium: Download from Remote as Git Patch). Either way, the authoritative copy lives in your repo.
From the Blog

Thinking out loud about test automation.

All articles →
Engineering March 2026 · 8 min read

Alliedium: The First Artifact-Based Test Automation Environment

Traditional automation tools produce logs. Alliedium produces artifacts — test plans, annotated screenshots, execution videos, and validation reports that let you understand, verify, and trust what the AI agent did.

Read article →
About Us

Engineers who've lived the QA problem.

Alliedium was founded by a team of test automation specialists who spent years solving the same problem across dozens of enterprise projects: the gap between manual test cases and reliable, maintainable automation is too wide, too slow to cross, and too dependent on a shrinking pool of automation engineers.

We built AutoMagic — and the cloud infrastructure around it — to close that gap entirely. Not by replacing testers, but by giving them a tool that handles the hard parts: turning plain English into structured, runnable tests; keeping those tests stable as the UI evolves; and running them on a fully managed cloud lab so teams don't have to maintain their own QA infrastructure.

Today we serve QA teams at companies of all sizes — from fast-growing software vendors to enterprise organisations with complex desktop applications that no browser-based testing tool can touch.

🔬
Deep automation expertise
Our team has been building test automation tooling for over two decades — including drivers for Windows, Java Swing, Oracle Forms, and web platforms that enterprise teams actually rely on.
☁️
Cloud-first, security-conscious
Every customer gets their own isolated cloud environment. Your code, your test data, and your LLM usage stay in your environment. We never share infrastructure between customers.
🤝
We do the setup, you get the results
Onboarding, deployment, configuration, and ongoing tuning of agents and drivers — that's our job. Your team focuses on testing, not infrastructure.
By the numbers
20+
Years in test automation
95%
Element locator stability over 2 years
5 + 1
Platforms GA + 1 beta
4
AI agents in the pipeline
~20 min
to create a 100-step test
1
Dedicated VPC per customer
Want to see it work on your app?
We'll run a live demo using one of your actual test cases — you'll see the full flow from plain English to a running automated test in under 30 minutes.
Request a Demo →