# Sorify

> AI-first browser testing and automation platform with MCP server, Chrome extension recorder, Playwright execution, scheduler, and Claude/Codex agent integration.

Sorify is an open-source, AI-first browser testing and automation platform built by Rakuten Technology (rakutentech) and released under the Apache License 2.0. It combines a Playwright-powered test runner, an MCP server for low-token AI agent control, a Chrome extension for session recording, and a built-in scheduler — all manageable from a single dashboard or via AI agents like Claude and Codex. The project reached v2.1 as of September 2026 and is self-hostable via Docker.

## What It Is

Sorify positions itself as an "AI Agentic QA Platform" designed for teams that want AI agents to write, manage, and heal browser tests. Rather than requiring developers to hand-write Playwright scripts, Sorify exposes every test management action through an MCP server so that Claude, Codex, or any OpenAI-compatible agent can create suites, trigger runs, investigate failures, and restore code versions using natural language commands. The dashboard provides a parallel UI for the same operations, and a REST API covers the full feature surface as well.

## How the MCP and Agent Layer Works

The MCP server is the central integration point. It exposes low-token-consumption tools covering suites, test cases, runs, screenshots, cookies, variables, and user management. Claude Code users install Sorify as a plugin from the repository marketplace; Codex users add it via `codex plugin marketplace add`. Once configured, agents use slash commands (`/sorify:gateway`, `/sorify:generate`, `/sorify:recording`) or Codex skills (`$sorify:sorify-gateway`, `$sorify:sorify-generate`) to interact with the platform. A built-in "My AI Agent" web chat inside the dashboard also connects to the MCP layer, supports page-aware context, and allows resumable handoffs to local agents.

## Chrome Extension Recorder

The Chrome extension operates in hybrid mode: while a developer browses, it captures browser events (clicks, inputs, file uploads, iframes, tabs, network activity) and stores them locally. The recorded session is then picked up by the `sorify-recorder-mcp` MCP server, which feeds the event sequence to an AI model to generate Playwright test code. This approach gives the AI richer context than DOM exploration alone. The extension is licensed separately under the BSD 3-Clause License.

## Integrations and Webhooks

Sorify supports three categories of bi-directional webhook integration:
- **CI webhook (inbound):** Trigger Sorify runs from GitHub Actions or any HTTP client using a token-authenticated suite webhook URL.
- **HTTP webhook (outbound):** Fire arbitrary HTTP requests before or after a test run.
- **GitHub Actions dispatch (outbound):** Trigger a GitHub Actions workflow from Sorify pre- or post-run, injecting Sorify run context as workflow inputs.

GitHub Apps (including GHES and GHEC) are managed from the Admin panel and support both OAuth login and action dispatch. Notifications are delivered via Microsoft Teams and email, with inline screenshots, cooldown controls, and digest summaries.

## Deployment Model

Sorify is self-hosted and ships with Docker Compose for the full stack (PHP/Laravel backend, MySQL). An ephemeral mode (introduced in v2.0) supports scalable, stateless deployments. The admin panel handles user management, role-based permissions, per-suite member access, and GitHub App registration. Internationalization covers English, Japanese, Simplified Chinese, and Malay.

## Update: v2.1 (September 2026)

The latest release, v2.1 (published 2026-09-13), adds the "My AI Agent" in-dashboard chat with OpenAI-compatible model profiles, page-aware context, MCP tool handoff, and AI model attribution on generated test code. Prior notable releases include v2.0 (ephemeral mode), v1.9 (Chrome extension memory leak fix, multi-connector pre/post webhooks), v1.8 (MCP run support, CI webhook improvements), v1.7 (cookie and variable support), and v1.6 (i18n expansion, suite/test duplication, OAuth login). The changelog is maintained in the repository README.

## Features
- AI test generation via Claude and Codex
- Playwright-powered browser test execution
- MCP server with low-token consumption
- Chrome extension for session recording
- Built-in scheduler with cron support
- Bi-directional webhook integrations (CI, HTTP, GitHub Actions)
- AI healing and failure investigation
- Test suite and test case management
- Bulk test creation and deletion
- Code version history and restore
- Cookie injection and pre-authenticated sessions
- Suite variables and proxy routing
- Role-based access control and per-suite members
- Microsoft Teams and email notifications with inline screenshots
- In-dashboard AI agent chat with MCP handoff
- Ephemeral self-hosting mode
- Docker deployment
- Admin panel with user management
- i18n support (English, Japanese, Chinese, Malay)
- Dark mode
- Run history, logs, and pass-rate analytics
- GitHub App integration for OAuth and Actions dispatch

## Integrations
Claude Code, OpenAI Codex, Playwright, GitHub Actions, GitHub Apps, GHES, GHEC, Microsoft Teams, Chrome, Docker, OpenAI-compatible APIs

## Platforms
WEB, API, CLI, BROWSER_EXTENSION

## Pricing
Open Source

## Version
v2.1

## Links
- Website: https://github.com/rakutentech/sorify
- Documentation: https://github.com/rakutentech/sorify/blob/master/INSTALL.md
- Repository: https://github.com/rakutentech/sorify
- EveryDev.ai: https://www.everydev.ai/tools/sorify