Jango

Product Introduction

1. Definition: Jango is a macOS-native desktop application for automated, multi-user acceptance and exploratory testing of web applications. It functions as an orchestration layer that deploys multiple AI-driven autonomous agents, each controlling an isolated Chromium browser instance via Playwright, to simulate real user interactions.
2. Core Value Proposition: It exists to solve the logistical bottleneck and high cost of coordinating human testers for multi-user application flows. Jango enables solo developers and small teams to automatically test complex, stateful interactions—like social feeds, marketplaces, order books, and collaborative tools—by providing a simulated environment with persistent, goal-oriented AI participants.

Main Features

1. Multi-Agent Browser Orchestration: Jango creates and manages a "cast" of independent AI participants. Each agent operates within its own fully isolated browser context, complete with separate cookies, local storage, and session state. This is powered by a bundled Node.js and Playwright environment, eliminating setup overhead. The agents navigate, authenticate using pre-configured test accounts, fill forms, click buttons, and perform sequences of actions based on natural language goals.
2. Persistent App Memory & Context: The platform maintains a structured memory system for each testing project. This includes encrypted login states for test accounts, a history of observed and actionable UI controls (e.g., "Post button," "Order submit form"), and evidentiary logs of all actions, screenshots, and errors from each test run. This creates repeatable test scenarios and reusable navigation hints, reducing test flakiness.
3. Live Session Control & Integration: Users can observe all agent browsers in real-time, pause execution, issue new directives mid-session, or manually take control of any agent's browser. Jango offers multiple entry points: a graphical dashboard, a CLI tool for CI/CD pipeline integration, and an MCP (Model Context Protocol) server for AI coding assistants, ensuring it fits into various development workflows.

Problems Solved

1. Pain Point: The extreme difficulty and resource cost of manually testing real-time, multi-user interactions such as chat sequencing, marketplace order matching, or collaborative document editing. Traditional methods require recruiting, scheduling, and briefing multiple human testers, which is slow, expensive, and non-repeatable.
2. Target Audience: Primary users are solo full-stack developers, startup engineering teams, and QA engineers building social platforms, SaaS collaboration tools, fintech sandboxes, or peer-to-peer marketplaces. They are typically resource-constrained and need to validate core interactive workflows before involving real users.
3. Use Cases: Essential for validating a new group chat feature where message order and delivery status must be correct across users; stress-testing a sandbox trading exchange with simultaneous buyers and sellers; ensuring role-based permissions work correctly in a project management tool; and automating user onboarding sequences that involve multiple accounts interacting.

Unique Advantages

1. Differentiation: Unlike single-browser end-to-end testing frameworks (e.g., Cypress, Playwright Test), Jango is built for concurrent multi-browser simulation. Unlike load testing tools (e.g., k6), it focuses on functional correctness and user journey simulation, not just system metrics. It differs from crowdsourced testing platforms by providing immediate, private, and repeatable automated sessions.
2. Key Innovation: Its integration of LLM-driven autonomous agents with precise browser automation control. The "AI participant" is not just a chatbot; it's an agent that can visually parse a live DOM, reason about goals, and perform precise UI interactions. The portable "cast" concept, where agent personas, goals, and app memory are saved and versioned, is a novel approach to test fixture management.

Frequently Asked Questions (FAQ)

1. How does Jango handle authentication and test user data? Jango requires you to provision separate test accounts within your application. It securely stores encrypted credentials using the macOS keychain and manages the login state for each AI participant in its isolated browser session, allowing for authentic multi-account testing.
2. Can I use Jango with any AI model or provider? Yes, Jango supports a Bring-Your-Own-Key (BYOK) model, allowing integration with OpenAI, Anthropic, or via the Vercel AI Gateway. Alternatively, you can purchase managed AI credits directly through Jango, which abstracts the provider complexity.
3. Is Jango suitable for testing production applications? No, Jango is designed explicitly for development and staging environments. It is intended to interact with your development URL and pre-configured test accounts to exercise features in a controlled, safe sandbox before deployment.
4. What types of web applications are not compatible with Jango? Applications relying heavily on canvas-based rendering (e.g., complex games), those requiring multi-factor authentication via external apps, or sites with CAPTCHA challenges may not be fully automatable by Jango's AI participants, as they depend on interpreting standard HTML controls.
5. Where is my application data and testing activity processed? Browser execution and page rendering occur locally on your Mac. Project metadata, action logs, and screenshots (evidence) are stored in Jango's cloud. Text from pages, along with agent goals and memories, is sent to your configured AI provider to generate actions.