Open-Source Testing · takahirom

arbigent

Arbigent is an open-source AI agent testing framework for iOS, Android, Web, and TV apps that breaks complex test scenarios into smaller, dependent tasks. It combines a UI-based scenario designer for QA engineers with a code interface for developers, and supports multiple AI providers (OpenAI, Gemini) to reduce vendor lock-in.

Talk to DEV.co Discuss Implementation

Source: GitHub — github.com/takahirom/arbigent

615

GitHub stars

Forks

Kotlin

Primary language

Apache-2.0

License (OSI-approved)

Key facts

Objective fields from the source. Values we can't verify are shown as “Unknown” rather than guessed.

Field	Value
Repository	takahirom/arbigent
Owner	takahirom
Primary language	Kotlin
License	Apache-2.0 — OSI-approved
Stars	615
Forks	61
Open issues	25
Latest release	0.74.0 (2026-06-25)
Last updated	2026-07-07
Source	https://github.com/takahirom/arbigent

What arbigent is

Written in Kotlin, Arbigent orchestrates AI-driven UI testing via scenario decomposition, UI tree optimization, and interceptor-based customization patterns. It integrates Maestro YAML flows for setup, supports Model Context Protocol (MCP) for external tool access, and includes AI-powered image assertions and stuck-screen detection for reliability.

Quickstart

Get the arbigent source

Clone the repository and explore it locally.

terminalbash

1git clone https://github.com/takahirom/arbigent.git2cd arbigent3# follow the project's README for install & configuration

Need it deployed, integrated, or customized instead? DEV.co ships production installs.

Best use cases

Complex multi-step mobile app workflows

Testing login → search → purchase flows where traditional brittle selectors fail. Arbigent breaks scenarios into dependent steps with AI re-evaluation, reducing flakiness from UI changes or dynamic content.

Cross-platform QA with minimal test code duplication

Test iOS, Android, and Web with one scenario framework. Non-programmers define tests in the UI; developers execute them programmatically, eliminating separate test suites per platform.

AI agent testing with provider flexibility

Organizations using internal AI models or wanting to avoid OpenAI/Gemini lock-in can implement custom AI providers via interceptor interfaces, then reuse all scenario definitions and orchestration logic.

Implementation considerations

AI provider API costs and rate limits will compound as test suites scale; cost-conscious teams should evaluate gpt-4o-mini vs. default gpt-4.1 models upfront.
Scenario decomposition is powerful but requires domain knowledge to partition tasks correctly; poorly designed scenarios may still flake or behave unpredictably.
MCP server integration adds flexibility but introduces external dependencies and potential security surface; each MCP server requires vetting.
UI tree optimization and AI hints (via accessibility labels) require app-side instrumentation; teams must coordinate with mobile developers to embed hints.
Stuck-screen detection and image-based assertions add latency; test execution time per scenario should be profiled in your environment.

When to avoid it — and what to weigh

You require enterprise SLA/support contracts — Arbigent is community-driven open source. No vendor guarantees, SLAs, or official support channels are documented. Suitable for self-supported teams only.
You need pixel-perfect UI regression testing — Arbigent is purpose-built for behavioral, task-driven testing, not layout validation. Traditional screenshot comparison or CSS regression tools are better suited.
You operate in highly regulated industries without code-review maturity — AI agent testing introduces non-determinism and external API calls (to OpenAI, Gemini, MCP servers). Audit trails and compliance validation are not evident in the documentation.
Your team has no Kotlin or mobile testing expertise — While the UI supports scenario design, integration, customization, and debugging require Kotlin knowledge. Learning curve is non-trivial for web-only teams.

License & commercial use

Apache License 2.0 (Apache-2.0). Permissive OSI license allowing commercial use, modification, and distribution with attribution and liability disclaimers.

Apache-2.0 explicitly permits commercial use, modification, and distribution. No royalties or licensing fees. However, this is the framework license only; costs are incurred by API calls to OpenAI, Gemini, and any hosted MCP servers. Teams should assess total cost of ownership (model API costs, infrastructure, maintenance) separately.

DEV.co evaluation signals

Editorial assessment — not user reviews. Directional, with an explicit confidence level.

Signal	Assessment
Maintenance	Active
Documentation	Adequate
License clarity	Clear
Deployment complexity	Moderate
DEV.co fit	Good
Assessment confidence	Medium

Security considerations

AI provider credentials (OpenAI, Gemini keys) must be managed securely; storing in YAML or environment variables risks exposure. MCP servers run with framework privileges and can access app state, logs, and external systems; untrusted MCP implementations could leak sensitive data. AI-generated actions (via external APIs) introduce non-deterministic behavior that audit/compliance teams may challenge. No documented security audit or vulnerability disclosure process.

Alternatives to consider

Maestro

Established declarative testing framework for mobile. YAML-driven, no AI, but faster and more predictable. Arbigent can integrate Maestro flows as setup; choose Maestro if you prefer determinism over AI adaptability.

Appium + AI (custom)

WebDriver-based testing with custom AI orchestration. Higher control but requires more engineering. Better for teams wanting to avoid vendor-specific frameworks.

Sauce Labs / BrowserStack

Managed cloud testing with some AI/ML features. Enterprise support and SLAs included. Trade-off: less customization, higher cost, vendor lock-in.

Software development agency

Build on arbigent with DEV.co software developers

Arbigent empowers QA and developers to build resilient AI-driven tests that adapt to UI changes. Start with a pilot scenario and measure flakiness reduction in your workflow.

Talk to DEV.co

Related open-source tools

Surfaced by semantic similarity across the DEV.co open-source index.

kagent AndroidViewClient koog

Related on DEV.co

Explore the category and the services that help you build with it.

Open-Source Testing AI Development Mobile App Development Talk to DEV.co

arbigent FAQ

Can I use Arbigent with my internal AI model instead of OpenAI/Gemini?

Yes. The interceptor pattern allows custom AI provider implementations. You would implement the provider interface and swap it in; scenarios and orchestration remain unchanged. Requires Kotlin development.

How much will AI API calls cost?

Depends on model choice (gpt-4.1 vs. gpt-4o-mini), scenario complexity, and test volume. Not disclosed in docs. Recommend pilot with cost monitoring before scaling.

Does Arbigent replace my existing UI test suite?

Not necessarily. Arbigent excels at high-level behavioral testing and exploratory scenarios. Use it alongside unit/integration tests and deterministic UI tests (Maestro, Espresso) for comprehensive coverage.

Is there a managed/SaaS version of Arbigent?

Not evident from the documentation or GitHub. Arbigent is self-hosted open source. Deployment, scaling, and cost management are your responsibility.

Software development & web development with DEV.co

From first prototype to production, DEV.co delivers software development services around tools like arbigent. Our software development agency staffs experienced software developers and web developers for custom software development, web development, integrations, and ongoing support across open-source testing and beyond.

Ready to Shift Mobile Testing Left?

Arbigent empowers QA and developers to build resilient AI-driven tests that adapt to UI changes. Start with a pilot scenario and measure flakiness reduction in your workflow.

Talk to DEV.co Schedule a Consultation