What platforms does Alumnium support?

Alumnium is available on Web App, iOS, Android, MacOS, Windows and Linux.

What are the limitations of Alumnium?

Benchmark and token-saving figures are published by the vendor and not independently confirmed here. Requires an AI provider account or local model, so model usage costs sit outside the tool. Natural-language steps can behave less deterministically than plain scripted assertions.

Alumnium

Name: Alumnium
Author: Alumnium

Open-source AI testing for coding agents and test engineers

Intermediate API

Visit Alumnium Back to AI tools

Screenshot of Alumnium, Open-source AI testing for coding agents and test engineers

What is Alumnium?

Alumnium is an open-source, MIT-licensed tool for AI-driven end-to-end testing of web, iOS and Android apps. It works as an MCP server for coding agents or as Java, Python and TypeScript clients layered on Selenium, Playwright, Appium or Maestro.

Alumnium is an open-source toolkit for end-to-end testing of web and mobile applications with AI. It ships in two forms: an MCP server that a coding agent can drive, and client libraries for Java, Python and TypeScript that sit inside test suites an engineering team already runs. The project is released under the MIT License and is developed in public on GitHub, with an active Discord community. For agent users, the setup is a single command that registers Alumnium as an MCP server with a supported coding agent such as Claude Code, Codex or Grok Build. The user then describes a flow and the expected outcome in plain language, and the agent uses Alumnium to operate and verify the application. A successful scenario can be saved as Markdown and replayed later, locally or in CI, through the Alumnium test runner. For engineers, the client wraps an existing Selenium, Playwright or Appium driver. Natural-language calls such as do, act and check are added next to ordinary framework code, so AI can be introduced one step at a time instead of replacing a stack. The site stresses efficiency. Alumnium keeps browser and mobile detail out of the agent's context and returns short summaries of what changed, which the project says cuts token use from 240K to 45K on a comparison task and costs about a quarter of Playwright. It is built to work with smaller, cheaper models, and supports hosted models from Anthropic, OpenAI, Google, xAI, Meta, DeepSeek and Mistral, plus local models through Ollama. The same intent can target web, iOS and Android, using Selenium or Playwright for web and Appium or Maestro for mobile. In the testing landscape, Alumnium sits between classic scripted frameworks and fully autonomous browser agents. It suits QA engineers and developers who want AI-assisted tests while keeping their existing tooling. The vendor also reports a 98.5% pass rate on the WebVoyager benchmark when paired with Claude Code, a self-published result worth verifying against your own applications.

How do you use Alumnium?

1Install the Alumnium CLI or client
For agents, install the binary with the install script for macOS and Linux or the PowerShell script for Windows. For engineers, add the alumnium package through npm, pip, Gradle or Maven.
2Choose an AI provider
Pick a supported model provider and set its credentials where required, such as an API key in the environment. Local models through Ollama are also an option.
3Connect the MCP server to your agent
Register Alumnium with an MCP-compatible coding agent, for example with the claude mcp add command pointing at alumnium mcp. The agent can then drive the app through Alumnium.
4Initialize the client in your tests
Engineers create an Alumni object around their existing Selenium, Playwright or Appium driver, and call quit during teardown.
5Write natural-language steps and checks
Add do or act calls for actions and check calls for assertions beside regular framework code, then run the suite as usual.
6Save and replay Markdown scenarios
Ask the coding agent to save a passing scenario as Markdown, then rerun it locally or in CI with the Alumnium test runner.

Pros and cons

Pros

Open source under the MIT License, with code and issues public on GitHubAI
Works through MCP with coding agents or as client libraries in existing test suitesAI
Adds to Selenium, Playwright, Appium or Maestro rather than replacing themAI
Supports many hosted model providers and local models via OllamaAI
Designed to save context and tokens, and to work with smaller, cheaper modelsAI

Cons

Benchmark and token-saving figures are published by the vendor and not independently confirmed hereAI
Requires an AI provider account or local model, so model usage costs sit outside the toolAI
Natural-language steps can behave less deterministically than plain scripted assertionsAI
Setup assumes comfort with package managers, test frameworks and CLI configurationAI
Homepage gives no hosted plan, pricing tiers or dedicated support channel beyond community and docsAI

How much does Alumnium cost?

Pricing

The homepage presents Alumnium as open source under the MIT License and lists no paid plans or prices. Costs from the chosen AI model provider are separate.

Learn more

Support

Documentation covers installation, configuration and MCP use, and there is a community Discord server plus a public GitHub repository. No dedicated support plan is described.

Learn more

Integrations

Works with Claude Code, Codex, Grok Build, Cursor, opencode and other MCP coding agents, plus Selenium, Playwright, Appium and Maestro. Models from Anthropic, OpenAI, Google, xAI, Meta, DeepSeek, Mistral, or local via Ollama.

Learn more

Features

MCP server for coding agents, Java, Python and TypeScript clients, natural-language do, act and check steps, Markdown scenarios with a test runner, web testing via Selenium or Playwright, iOS and Android via Appium or Maestro, multi-session support, and many hosted or local model options.

Frequently asked questions about Alumnium

How much does Alumnium cost?
The homepage presents Alumnium as open source under the MIT License and lists no paid plans or prices. Costs from the chosen AI model provider are separate.
Does Alumnium have an API?
Yes, Alumnium offers an API.
How do you use Alumnium?
The walkthrough on this page covers 6 steps: 1. Install the Alumnium CLI or client 2. Choose an AI provider 3. Connect the MCP server to your agent 4. Initialize the client in your tests 5. Write natural-language steps and checks 6. Save and replay Markdown scenarios.
What platforms does Alumnium support?
Alumnium is available on Web App, iOS, Android, MacOS, Windows and Linux.
What does Alumnium integrate with?
Works with Claude Code, Codex, Grok Build, Cursor, opencode and other MCP coding agents, plus Selenium, Playwright, Appium and Maestro. Models from Anthropic, OpenAI, Google, xAI, Meta, DeepSeek, Mistral, or local via Ollama.
What are the limitations of Alumnium?
Benchmark and token-saving figures are published by the vendor and not independently confirmed here. Requires an AI provider account or local model, so model usage costs sit outside the tool. Natural-language steps can behave less deterministically than plain scripted assertions.