reverse-api-engineer 0.2.8__tar.gz → 0.2.10__tar.gz

{reverse_api_engineer-0.2.8 → reverse_api_engineer-0.2.10}/.claude/settings.local.json

@@ -51,7 +51,26 @@
       "WebFetch(domain:localhost)",
       "Bash(curl:*)",
       "WebFetch(domain:docs.litellm.ai)",
-      "Bash(git reset:*)"
+      "Bash(git reset:*)",
+      "Bash(ls:*)",
+      "Bash(if [ -n \"$PYPI_TOKEN\" ])",
+      "Bash(then)",
+      "Bash(elif [ -n \"$UV_PUBLISH_TOKEN\" ])",
+      "Bash(else)",
+      "Bash(fi)",
+      "Bash(python -m pytest:*)",
+      "WebFetch(domain:opencode.ai)",
+      "Skill(plugin-dev:plugin-structure)",
+      "Bash(tree:*)",
+      "Bash(mkdir:*)",
+      "Skill(plugin-dev:command-development)",
+      "mcp__plugin_reverse-api-engineer_rae-playwright-mcp__browser_navigate",
+      "mcp__plugin_reverse-api-engineer_rae-playwright-mcp__browser_close",
+      "mcp__plugin_reverse-api-engineer_rae-playwright-mcp__browser_click",
+      "mcp__plugin_reverse-api-engineer_rae-playwright-mcp__browser_snapshot",
+      "mcp__plugin_reverse-api-engineer_rae-playwright-mcp__browser_run_code",
+      "mcp__plugin_reverse-api-engineer_rae-playwright-mcp__browser_network_requests",
+      "Skill(reverse-api-engineer:reverse-engineering-api)"
     ]
   }
 }

reverse_api_engineer-0.2.10/.claude-plugin/marketplace.json

@@ -0,0 +1,14 @@
+{
+  "name": "reverse-api-engineer",
+  "owner": {
+    "name": "Kalil Sama Bouzigues",
+    "email": "kalil.bouzigues@gmail.com"
+  },
+  "plugins": [
+    {
+      "name": "reverse-api-engineer",
+      "source": "./plugins/reverse-api-engineer",
+      "description": "A plugin to reverse engineer APIs from websites"
+    }
+  ]
+}

{reverse_api_engineer-0.2.8 → reverse_api_engineer-0.2.10}/.gitignore

@@ -88,4 +88,6 @@ RELEASE_NOTES.md
 RELEASE.md
 prompt*.md
 
-test*
+test*
+
+FEATURES.md

{reverse_api_engineer-0.2.8 → reverse_api_engineer-0.2.10}/CHANGELOG.md

@@ -5,6 +5,57 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [0.2.10] - 2026-01-03
+
+### Added
+- **Claude Code Plugin**: Official plugin for seamless integration with Claude Code CLI
+  - Three operation modes: manual browser capture, autonomous agent browsing, and re-engineering from HAR files
+  - Comprehensive skill system with progressive disclosure of reverse engineering techniques
+  - Slash commands for quick access: `/agent`, `/engineer`, `/manual`
+  - Reference documentation for HAR analysis and authentication patterns
+  - API client templates for common patterns
+- **Example API Clients**: Added production-ready examples for major platforms
+  - Apple Jobs API client with field extraction utilities
+  - Ashby Jobs API client with comprehensive endpoint coverage
+  - Ikea API client for product search and catalog browsing
+  - Uber Careers API client with pagination support
+- **Engineer Tagging System**: Enhanced metadata tracking for generated API clients
+  - Automatic tagging of runs with descriptive identifiers
+  - Improved organization and searchability of reverse-engineered APIs
+
+### Changed
+- **Enhanced Auto Mode**: Improved MCP browser integration with better error handling
+- **Better Sync Fallback**: More robust file synchronization with fallback mechanisms
+- **Code Quality**: Comprehensive formatting and linting improvements across codebase
+
+### Fixed
+- **Agent Mode Screenshots**: Reduced unnecessary screenshot captures in agent mode
+- **Sync Error Handling**: Fixed sync fallback when primary sync method fails
+- **Path Handling**: Corrected CLAUDE.md documentation paths
+- **Import Errors**: Fixed missing imports in various modules
+
+## [0.2.9] - 2025-12-30
+
+### Added
+- **Real-time File Sync**: Watch and automatically sync generated scripts to local directory
+  - Debounced file watching with configurable delay (default 500ms)
+  - Visual feedback for sync operations in terminal UI
+  - Prevents overwriting existing directories by appending counter suffix
+- **MCP Browser Integration**: Native integration with `rae-playwright-mcp` for auto mode
+  - Seamless browser automation via Model Context Protocol
+  - Works with both Claude SDK and OpenCode SDK
+  - Combines browser control and real-time reverse engineering in single workflow
+- **CLAUDE.md Autogeneration**: Automatic generation of project documentation for Claude Code
+
+### Changed
+- **Enhanced Settings Management**: Improved settings configuration and UI
+- **Better Sync Error Handling**: Improved error handling and resource cleanup for sync operations
+
+### Fixed
+- **Sync Directory Overwrite**: Fixed issue where sync would overwrite existing directories
+- **Sync Resource Leaks**: Fixed memory leaks when sync errors occurred
+- **UI Improvements**: Various UI fixes and enhancements
+
 ## [0.2.8] - 2025-12-28
 
 ### Added

reverse_api_engineer-0.2.10/CLAUDE.md

@@ -0,0 +1,220 @@
+# CLAUDE.md
+
+This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
+
+## Project Overview
+
+Reverse API Engineer is a CLI tool that captures browser traffic (HAR files) and uses AI to automatically generate production-ready Python API clients. It supports three modes: manual browser capture, autonomous AI agent browsing, and re-engineering from previous captures.
+
+## Core Architecture
+
+### SDK Dispatch Pattern
+The project uses a **dual-SDK architecture** controlled by `config.json`:
+
+- **Claude SDK** (`engineer.py` → `ClaudeEngineer`): Uses `claude-agent-sdk` for direct Claude API integration
+- **OpenCode SDK** (`opencode_engineer.py` → `OpenCodeEngineer`): Uses httpx to communicate with local OpenCode server (http://127.0.0.1:4096)
+
+Both inherit from `BaseEngineer` (abstract base class) which defines the common interface and shared prompt building logic. The SDK selection is determined by the `sdk` field in config.
+
+### Key Components
+
+- **CLI Layer** (`cli.py`): Main entry point with mode cycling (manual/engineer/agent), slash commands, and interactive prompts
+- **Browser** (`browser.py`): Playwright-based HAR recording with stealth mode and anti-detection measures
+  - `ManualBrowser`: User-controlled browser with HAR capture
+  - `run_agent_browser()`: Autonomous agent browsing (browser-use or stagehand)
+- **Configuration** (`config.py`): JSON-based config manager at `~/.reverse-api/config.json`
+- **Session Management** (`session.py`): Tracks run history with costs and metadata
+- **Message Store** (`messages.py`): Persists full conversation logs per run
+- **UI** (`tui.py`, `opencode_ui.py`): Rich-based terminal UI for progress tracking
+
+### Data Flow
+1. Browser captures HAR → saved to `~/.reverse-api/runs/har/{run_id}`
+1. Browser captures HAR → saved to `~/.reverse-api/runs/har/{run_id}`
+2. Engineer analyzes HAR with LLM → generates Python scripts
+3. Scripts saved to:
+   - `~/.reverse-api/runs/scripts/{run_id}` (permanent)
+   - `~/.reverse-api/runs/scripts/{run_id}` (permanent)
+   - `./scripts/{descriptive_name}/` (local copy)
+
+## Development Commands
+
+### Setup
+```bash
+# Clone and install dependencies
+git clone https://github.com/kalil0321/reverse-api-engineer.git
+cd reverse-api-engineer
+uv sync
+playwright install chromium
+
+# Run locally (development mode)
+uv run reverse-api-engineer
+```
+
+### Building
+```bash
+# Clean build (removes all caches and artifacts)
+./scripts/clean_build.sh
+
+# Manual build
+uv build
+```
+
+### Testing
+```bash
+# Test as installed package (creates isolated venv)
+python -m venv test_env
+test_env/bin/pip install dist/*.whl
+test_env/bin/reverse-api-engineer
+
+# Test with uv tool
+uv tool install .
+reverse-api-engineer
+```
+
+### Linting & Static Analysis
+```bash
+# Run all checks
+./scripts/lint.sh
+
+# Run specific tools
+uv run ruff check src     # Linter
+uv run ruff format src    # Formatter
+uv run mypy src           # Static analysis
+```
+
+## Release Process
+
+**Single source of truth: `pyproject.toml`**
+
+1. Update version in `pyproject.toml` only
+2. Update `CHANGELOG.md`
+3. Run `./scripts/clean_build.sh`
+4. Commit and tag:
+   ```bash
+   git add pyproject.toml CHANGELOG.md
+   git commit -m "release: vX.Y.Z - description"
+   git tag vX.Y.Z
+   git push origin main
+   git push origin vX.Y.Z
+   ```
+5. Publish to PyPI:
+   ```bash
+   source .env  # contains UV_PUBLISH_TOKEN or PYPI_TOKEN
+   uv publish
+   ```
+
+**Important:** Never edit `src/reverse_api/__init__.py` for versioning—it reads from package metadata via `importlib.metadata`.
+
+## Configuration System
+
+### Config Location
+`~/.reverse-api/config.json` with these fields:
+
+```json
+{
+  "claude_code_model": "claude-sonnet-4-5",    // For Claude SDK
+  "opencode_provider": "anthropic",            // For OpenCode SDK
+  "opencode_model": "claude-sonnet-4-5",       // For OpenCode SDK
+  "sdk": "claude",                             // "claude" or "opencode"
+  "agent_provider": "browser-use",             // "browser-use" or "stagehand"
+  "browser_use_model": "bu-llm",               // Browser-use agent model
+  "stagehand_model": "openai/computer-use-preview-2025-03-11",
+  "output_dir": null                           // Custom output dir (null = use ~/.reverse-api/runs)
+}
+```
+
+### Model Naming
+- **Claude models**: `claude-sonnet-4-5`, `claude-opus-4-5`, `claude-haiku-4-5`
+- **Gemini models**: `gemini-3-flash`, `gemini-3-pro`, `gemini-3-pro-low`, `gemini-3-pro-high`
+- **Antigravity models** (free tier): Require OpenCode SDK with `opencode_provider: "google"`
+- **Agent models**: Format varies by provider
+  - Browser-use: `bu-llm` or `{provider}/{model}` (e.g., `openai/gpt-4`)
+  - Stagehand: `{provider}/{model}` (e.g., `openai/computer-use-preview-2025-03-11`)
+
+## Pricing System
+
+### Three-tier fallback for cost calculation:
+1. **Local pricing** (`pricing.py`): Built-in prices for Claude/Gemini models
+2. **LiteLLM pricing** (optional `[pricing]` extra): Extended coverage for 100+ models
+3. **Default pricing**: Falls back to Claude Sonnet 4.5 pricing
+
+Pricing tracks:
+- Input/output tokens
+- Cache creation/read tokens
+- Reasoning tokens (for thinking models)
+
+## Agent Mode
+
+### Requirements
+Agent mode requires browser-use from specific git commit (has HAR recording support):
+```bash
+uv tool install 'reverse-api-engineer[agent]' --with 'browser-use @ git+https://github.com/browser-use/browser-use.git@49a345fb19e9f12befc5cc1658e0033873892455'
+```
+
+### Agent Providers
+- **browser-use**: Multi-model support (Browser-Use LLM, OpenAI, Google)
+- **stagehand**: Computer Use models only (OpenAI, Anthropic)
+
+API keys via environment variables: `BROWSER_USE_API_KEY`, `OPENAI_API_KEY`, `ANTHROPIC_API_KEY`, `GOOGLE_API_KEY`
+
+## Code Style Conventions
+
+- **Type hints**: Use throughout (`from typing import Optional, Dict, Any`)
+- **Async/await**: Engineers use async for LLM streaming
+- **Error handling**: Wrap external calls in try-except with user-friendly messages
+- **UI consistency**: Use Rich for all terminal output (no plain print statements)
+- **Config migration**: `config.py` includes backward compatibility logic for renamed keys
+
+## Common Tasks
+
+### Adding a new model
+1. Add pricing to `pricing.py` MODEL_PRICING dict (if not in LiteLLM)
+2. Update model choices in `tui.py` get_model_choices() if needed
+3. Update README.md model list
+
+### Adding a new agent provider
+1. Add provider logic to `browser.py` run_agent_browser()
+2. Add config fields to `config.py` DEFAULT_CONFIG
+3. Update settings menu in `cli.py`
+4. Document in README.md
+
+### Debugging
+- **OpenCode**: Set `OPENCODE_DEBUG=1` for detailed logs
+- **Stagehand logs**: Suppressed by `_suppress_stagehand_logs()` in `browser.py`
+- **Message logs**: All LLM interactions saved to `~/.reverse-api/runs/messages/{run_id}.jsonl`
+
+## File Structure Notes
+
+```
+src/reverse_api/
+├── cli.py              # Main entry point, mode cycling, slash commands
+├── base_engineer.py    # Abstract base for SDK implementations
+├── engineer.py         # Claude SDK implementation
+├── opencode_engineer.py # OpenCode SDK implementation
+├── browser.py          # Playwright HAR capture + agent mode
+├── config.py           # Configuration manager
+├── session.py          # Run history and metadata
+├── messages.py         # Message persistence
+├── pricing.py          # Model pricing database
+├── tui.py              # Terminal UI (Rich-based)
+├── opencode_ui.py      # OpenCode-specific UI extensions
+└── utils.py            # Path helpers, run ID generation
+
+~/.reverse-api/
+├── config.json         # User configuration
+├── history.json        # Run history with costs
+└── runs/               # Organized by data type
+    ├── har/{run_id}/   # Captured traffic per run
+    ├── scripts/{run_id}/ # Generated API clients per run
+    └── messages/       # LLM conversation logs
+        └── {run_id}.jsonl
+```
+
+## Dependencies Management
+
+Uses `uv` for dependency management:
+- **Core deps**: playwright, claude-agent-sdk, rich, questionary, anthropic
+- **Optional [agent]**: stagehand (browser-use must be installed separately from git)
+- **Optional [pricing]**: litellm (for extended model pricing)
+
+Always use `uv sync` after pulling changes.

reverse_api_engineer-0.2.10/INTERVIEW.md

@@ -0,0 +1,5 @@
+Read this @SPEC.md and interview me in detail using the AskUserQuestionTool about 
+literally anything: technical implementation, UI & UX, concerns, tradeoffs, etc. 
+but make sure the questions are not obvious.
+
+Be very in-depth and continue interviewing me continually until it’s complete, then write the spec to the file.

{reverse_api_engineer-0.2.8 → reverse_api_engineer-0.2.10}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: reverse-api-engineer
-Version: 0.2.8
+Version: 0.2.10
 Summary: A tool to capture browser traffic for API reverse engineering
 Project-URL: Homepage, https://github.com/kalil0321/reverse-api-engineer
 Project-URL: Repository, https://github.com/kalil0321/reverse-api-engineer
@@ -29,6 +29,7 @@ Requires-Dist: playwright>=1.40.0
 Requires-Dist: questionary>=2.0.0
 Requires-Dist: requests>=2.32.5
 Requires-Dist: rich>=13.0.0
+Requires-Dist: watchdog>=3.0.0
 Provides-Extra: agent
 Requires-Dist: stagehand; extra == 'agent'
 Provides-Extra: pricing
@@ -36,18 +37,16 @@ Requires-Dist: litellm>=1.0.0; extra == 'pricing'
 Description-Content-Type: text/markdown
 
 <div align="center">
-  <img src="assets/reverse-api-banner.svg" alt="Reverse API Engineer Banner">
+  <img src="https://raw.githubusercontent.com/kalil0321/reverse-api-engineer/main/assets/reverse-api-banner.svg" alt="Reverse API Engineer Banner">
 </div>
 
-# Reverse API Engineer
-
-[![PyPI version](https://badge.fury.io/py//reverse-api-engineer.svg)](https://badge.fury.io/py/reverse-api-engineer)
+![PyPI](https://img.shields.io/pypi/v/reverse-api-engineer)
 [![Python 3.11+](https://img.shields.io/badge/python-3.11+-blue.svg)](https://www.python.org/downloads/)
 [![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
 
 CLI tool that captures browser traffic and automatically generates production-ready Python API clients. No more manual reverse engineering—just browse, capture, and get clean API code.
 
-![Reverse API Engineer Demo](assets/reverse-api-engineer.gif)
+![Reverse API Engineer Demo](https://raw.githubusercontent.com/kalil0321/reverse-api-engineer/main/assets/reverse-api-engineer.gif)
 
 ## Table of Contents
 
@@ -60,10 +59,11 @@ CLI tool that captures browser traffic and automatically generates production-re
   - [Agent Mode](#agent-mode)
 - [Configuration](#-configuration)
   - [Model Selection](#model-selection)
-  - [💸 Free Models with Antigravity](#-free-model-options-with-antigravity)
+  - [💸 Free Models with Antigravity](#free-model-options-with-antigravity)
   - [Agent Configuration](#agent-configuration)
   - [SDK Selection](#sdk-selection)
 - [CLI Commands](#-cli-commands)
+- [Claude Code Plugin](#-claude-code-plugin)
 - [Examples](#-examples)
 - [Development](#-development)
 - [Contributing](#-contributing)
@@ -71,7 +71,7 @@ CLI tool that captures browser traffic and automatically generates production-re
 ## ✨ Features
 
 - 🌐 **Browser Automation**: Built on Playwright with stealth mode for realistic browsing
-- 🤖 **Autonomous Agent Mode**: Fully automated browser interaction using AI agents (browser-use, stagehand)
+- 🤖 **Autonomous Agent Mode**: Fully automated browser interaction using AI agents (auto mode with MCP, browser-use, stagehand)
 - 📊 **HAR Recording**: Captures all network traffic in HTTP Archive format
 - 🧠 **AI-Powered Generation**: Uses Claude 4.5 to analyze traffic and generate clean Python code
 - 🔌 **Multi-SDK Support**: Native integration with Claude and OpenCode SDKs
@@ -141,7 +141,7 @@ reverse-api-engineer
 The CLI has three modes (cycle with **Shift+Tab**):
 - **manual**: Browser capture + AI generation
 - **engineer**: Re-process existing captures
-- **agent**: Autonomous AI browser agent
+- **agent**: Autonomous AI browser agent (default: auto mode with MCP-based browser + real-time reverse engineering)
 
 Example workflow:
 ```bash
@@ -185,15 +185,20 @@ reverse-api-engineer engineer <run_id>
 
 Fully automated browser interaction using AI agents:
 
-1. Install with agent support (see [Installation](#-installation))
-2. Start CLI and switch to agent mode (Shift+Tab)
-3. Enter task description (e.g., "Click on the first job listing")
-4. Optionally provide starting URL
-5. Agent automatically navigates and interacts
-6. HAR captured automatically
-7. Optionally generate API client
+1. Start CLI and switch to agent mode (Shift+Tab)
+2. Enter task description (e.g., "Click on the first job listing")
+3. Optionally provide starting URL
+4. Agent automatically navigates and interacts
+5. HAR captured automatically
+6. API client generated automatically
+
+**Agent Provider Options:**
 
-**Note:** Requires browser-use from the specific git commit shown in installation instructions (includes HAR recording support).
+- **auto** (default): Uses MCP-based browser automation with Claude Agent SDK & Opencode. Combines browser control and real-time reverse engineering in a single workflow. No additional installation required beyond the base package.
+- **browser-use**: Uses browser-use library for browser automation. Requires installation with `[agent]` extra and browser-use from specific git commit (includes HAR recording support).
+- **stagehand**: Uses Stagehand for browser automation with Computer Use models.
+
+Change agent provider in `/settings` → "agent provider".
 
 ## 🔧 Configuration
 
@@ -204,7 +209,7 @@ Settings stored in `~/.reverse-api/config.json`:
   "opencode_provider": "anthropic",
   "opencode_model": "claude-sonnet-4-5",
   "sdk": "claude",
-  "agent_provider": "browser-use",
+  "agent_provider": "auto",
   "browser_use_model": "bu-llm",
   "stagehand_model": "openai/computer-use-preview-2025-03-11",
   "output_dir": null
@@ -223,6 +228,8 @@ Change in `/settings` or via CLI:
 reverse-api-engineer manual --model claude-sonnet-4-5
 ```
 
+If you use Opencode, look at the [models](https://models.dev). Some are free like `opencode/grok-code`).
+
 ### Free Model Options with Antigravity
 
 You can use free models via [Antigravity](https://github.com/NoeFabris/opencode-antigravity-auth) for API generation. See [available models](https://github.com/NoeFabris/opencode-antigravity-auth#available-models) for the full list.
@@ -252,7 +259,8 @@ You can use free models via [Antigravity](https://github.com/NoeFabris/opencode-
 Configure AI agents for autonomous browser automation.
 
 **Agent Providers:**
-- **browser-use** (default): Supports Browser-Use LLM, OpenAI, and Google models
+- **auto** (default): MCP-based browser automation with real-time reverse engineering. Uses Claude Agent SDK with browser MCP tools. Combines browser control and API reverse engineering in a single unified workflow. Works with Claude SDK (default) or OpenCode SDK.
+- **browser-use**: Supports Browser-Use LLM, OpenAI, and Google models. Requires installation with `[agent]` extra.
 - **stagehand**: Supports OpenAI and Anthropic Computer Use models
 
 **Agent Models:**
@@ -294,6 +302,18 @@ Use these slash commands while in the CLI:
 - `/help` - Show all commands
 - `/exit` - Quit
 
+## 🔌 Claude Code Plugin
+
+Install the plugin in [Claude Code](https://claude.com/claude-code):
+
+```bash
+claude # Open REPL
+/plugin marketplace add kalil0321/reverse-api-engineer
+/plugin install reverse-api-engineer@reverse-api-engineer
+```
+
+See [plugin documentation](plugins/reverse-api-engineer/README.md) for commands, agents, skills, and usage examples.
+
 ## 💡 Examples
 
 ### Example: Reverse Engineering a Job Board API

{reverse_api_engineer-0.2.8 → reverse_api_engineer-0.2.10}/README.md

@@ -1,16 +1,14 @@
 <div align="center">
-  <img src="assets/reverse-api-banner.svg" alt="Reverse API Engineer Banner">
+  <img src="https://raw.githubusercontent.com/kalil0321/reverse-api-engineer/main/assets/reverse-api-banner.svg" alt="Reverse API Engineer Banner">
 </div>
 
-# Reverse API Engineer
-
-[![PyPI version](https://badge.fury.io/py//reverse-api-engineer.svg)](https://badge.fury.io/py/reverse-api-engineer)
+![PyPI](https://img.shields.io/pypi/v/reverse-api-engineer)
 [![Python 3.11+](https://img.shields.io/badge/python-3.11+-blue.svg)](https://www.python.org/downloads/)
 [![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
 
 CLI tool that captures browser traffic and automatically generates production-ready Python API clients. No more manual reverse engineering—just browse, capture, and get clean API code.
 
-![Reverse API Engineer Demo](assets/reverse-api-engineer.gif)
+![Reverse API Engineer Demo](https://raw.githubusercontent.com/kalil0321/reverse-api-engineer/main/assets/reverse-api-engineer.gif)
 
 ## Table of Contents
 
@@ -23,10 +21,11 @@ CLI tool that captures browser traffic and automatically generates production-re
   - [Agent Mode](#agent-mode)
 - [Configuration](#-configuration)
   - [Model Selection](#model-selection)
-  - [💸 Free Models with Antigravity](#-free-model-options-with-antigravity)
+  - [💸 Free Models with Antigravity](#free-model-options-with-antigravity)
   - [Agent Configuration](#agent-configuration)
   - [SDK Selection](#sdk-selection)
 - [CLI Commands](#-cli-commands)
+- [Claude Code Plugin](#-claude-code-plugin)
 - [Examples](#-examples)
 - [Development](#-development)
 - [Contributing](#-contributing)
@@ -34,7 +33,7 @@ CLI tool that captures browser traffic and automatically generates production-re
 ## ✨ Features
 
 - 🌐 **Browser Automation**: Built on Playwright with stealth mode for realistic browsing
-- 🤖 **Autonomous Agent Mode**: Fully automated browser interaction using AI agents (browser-use, stagehand)
+- 🤖 **Autonomous Agent Mode**: Fully automated browser interaction using AI agents (auto mode with MCP, browser-use, stagehand)
 - 📊 **HAR Recording**: Captures all network traffic in HTTP Archive format
 - 🧠 **AI-Powered Generation**: Uses Claude 4.5 to analyze traffic and generate clean Python code
 - 🔌 **Multi-SDK Support**: Native integration with Claude and OpenCode SDKs
@@ -104,7 +103,7 @@ reverse-api-engineer
 The CLI has three modes (cycle with **Shift+Tab**):
 - **manual**: Browser capture + AI generation
 - **engineer**: Re-process existing captures
-- **agent**: Autonomous AI browser agent
+- **agent**: Autonomous AI browser agent (default: auto mode with MCP-based browser + real-time reverse engineering)
 
 Example workflow:
 ```bash
@@ -148,15 +147,20 @@ reverse-api-engineer engineer <run_id>
 
 Fully automated browser interaction using AI agents:
 
-1. Install with agent support (see [Installation](#-installation))
-2. Start CLI and switch to agent mode (Shift+Tab)
-3. Enter task description (e.g., "Click on the first job listing")
-4. Optionally provide starting URL
-5. Agent automatically navigates and interacts
-6. HAR captured automatically
-7. Optionally generate API client
+1. Start CLI and switch to agent mode (Shift+Tab)
+2. Enter task description (e.g., "Click on the first job listing")
+3. Optionally provide starting URL
+4. Agent automatically navigates and interacts
+5. HAR captured automatically
+6. API client generated automatically
+
+**Agent Provider Options:**
 
-**Note:** Requires browser-use from the specific git commit shown in installation instructions (includes HAR recording support).
+- **auto** (default): Uses MCP-based browser automation with Claude Agent SDK & Opencode. Combines browser control and real-time reverse engineering in a single workflow. No additional installation required beyond the base package.
+- **browser-use**: Uses browser-use library for browser automation. Requires installation with `[agent]` extra and browser-use from specific git commit (includes HAR recording support).
+- **stagehand**: Uses Stagehand for browser automation with Computer Use models.
+
+Change agent provider in `/settings` → "agent provider".
 
 ## 🔧 Configuration
 
@@ -167,7 +171,7 @@ Settings stored in `~/.reverse-api/config.json`:
   "opencode_provider": "anthropic",
   "opencode_model": "claude-sonnet-4-5",
   "sdk": "claude",
-  "agent_provider": "browser-use",
+  "agent_provider": "auto",
   "browser_use_model": "bu-llm",
   "stagehand_model": "openai/computer-use-preview-2025-03-11",
   "output_dir": null
@@ -186,6 +190,8 @@ Change in `/settings` or via CLI:
 reverse-api-engineer manual --model claude-sonnet-4-5
 ```
 
+If you use Opencode, look at the [models](https://models.dev). Some are free like `opencode/grok-code`).
+
 ### Free Model Options with Antigravity
 
 You can use free models via [Antigravity](https://github.com/NoeFabris/opencode-antigravity-auth) for API generation. See [available models](https://github.com/NoeFabris/opencode-antigravity-auth#available-models) for the full list.
@@ -215,7 +221,8 @@ You can use free models via [Antigravity](https://github.com/NoeFabris/opencode-
 Configure AI agents for autonomous browser automation.
 
 **Agent Providers:**
-- **browser-use** (default): Supports Browser-Use LLM, OpenAI, and Google models
+- **auto** (default): MCP-based browser automation with real-time reverse engineering. Uses Claude Agent SDK with browser MCP tools. Combines browser control and API reverse engineering in a single unified workflow. Works with Claude SDK (default) or OpenCode SDK.
+- **browser-use**: Supports Browser-Use LLM, OpenAI, and Google models. Requires installation with `[agent]` extra.
 - **stagehand**: Supports OpenAI and Anthropic Computer Use models
 
 **Agent Models:**
@@ -257,6 +264,18 @@ Use these slash commands while in the CLI:
 - `/help` - Show all commands
 - `/exit` - Quit
 
+## 🔌 Claude Code Plugin
+
+Install the plugin in [Claude Code](https://claude.com/claude-code):
+
+```bash
+claude # Open REPL
+/plugin marketplace add kalil0321/reverse-api-engineer
+/plugin install reverse-api-engineer@reverse-api-engineer
+```
+
+See [plugin documentation](plugins/reverse-api-engineer/README.md) for commands, agents, skills, and usage examples.
+
 ## 💡 Examples
 
 ### Example: Reverse Engineering a Job Board API