reverse-api-engineer 0.9.0

A tool to capture browser traffic for API reverse engineering

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.

Agent mode

Manual mode

Table of Contents

• Features
• Installation
• Quick Start
• Usage Modes
  • Manual Mode
  • Engineer Mode
  • Agent Mode
  • Collector Mode
• Configuration
  • Model Selection
  • Agent Configuration
  • SDK Selection
• CLI Commands
• Examples
• Development
• Contributing

✨ Features

• 🌐 Browser Automation: Built on Playwright with stealth mode for realistic browsing
• 🤖 Autonomous Agent Mode: Fully automated browser interaction using AI agents via MCP (Playwright MCP or Chrome DevTools MCP)
• 📊 HAR Recording: Captures all network traffic in HTTP Archive format
• 🧠 AI-Powered Generation: Uses Claude 4.6 to analyze traffic and generate clean Python code
• 🔍 Collector Mode: Data collection with automatic JSON/CSV export
• 🔌 Multi-SDK Support: Native integration with Claude and OpenCode SDKs
• 💻 Interactive CLI: Minimalist terminal interface with mode cycling (Shift+Tab)
• 📦 Production Ready: Generated scripts include error handling, type hints, and documentation
• 💾 Session History: All runs saved locally with full message logs
• 💰 Cost Tracking: Detailed token usage and cost estimation with cache support

Limitations

• This tool executes code locally using Claude Code—please monitor output
• Some websites employ advanced bot-detection that may limit capture or require manual interaction

🚀 Installation

Using uv (recommended)

uv tool install reverse-api-engineer

Using pip

pip install reverse-api-engineer

Post-installation

Install Playwright browsers:

playwright install chromium

Pricing

Reverse API Engineer includes built-in pricing for the most common models (Claude 4.6, Gemini 3, GPT-4.1/5). Cost tracking uses a 2-tier fallback:

1. Local pricing table — built-in pricing for common models
2. Default fallback — Claude Sonnet 4.6 pricing for unknown models

If litellm is independently installed in your environment, an extended pricing lookup is auto-detected for 100+ additional models. We no longer ship a [pricing] extra; install litellm directly if you need that coverage.

🚀 Quick Start

Launch the interactive CLI:

reverse-api-engineer

The CLI has four modes (cycle with Shift+Tab):

• manual: Browser capture + AI generation
• engineer: Re-process existing captures
• agent: Autonomous AI browser agent (default: auto mode with MCP-based browser + real-time reverse engineering)
• collector: AI-powered web data collection (very minimalist version for now)

Example workflow:

$ reverse-api-engineer
> fetch all apple jobs from their careers page

# Browser opens, navigate and interact
# Close browser when done
# AI generates production-ready API client

# Scripts saved to: ./scripts/apple_jobs_api/

📖 Usage Modes

Manual Mode

Full pipeline with manual browser interaction:

1. Start the CLI: reverse-api-engineer
2. Enter task description (e.g., "Fetch Apple job listings")
3. Optionally provide starting URL
4. Browse and interact with the website
5. Close browser when done
6. AI automatically generates the API client

Output locations:

• ~/.reverse-api/runs/scripts/{run_id}/ (permanent storage)
• ./scripts/{descriptive_name}/ (local copy with readable name)

Engineer Mode

Re-run AI generation on a previous capture:

# Switch to engineer mode (Shift+Tab) and enter run_id
# Or use command line:
reverse-api-engineer engineer <run_id>

Agent Mode

Fully automated browser interaction using AI agents:

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:

• auto (default): Uses Playwright MCP browser automation with Claude Agent SDK & Opencode. Combines browser control and real-time reverse engineering in a single workflow.
• chrome-mcp: Uses Chrome DevTools MCP to drive your real Chrome browser (with existing sessions, cookies, and auth). Requires Chrome 146+ and Node.js 20.19+.

Change agent provider in /settings → "agent provider".

Collector Mode

Web data collection using Claude Agent SDK:

1. Start CLI and switch to collector mode (Shift+Tab)
2. Enter a natural language prompt describing the data to collect (e.g., "Find 3 JS frameworks")
3. The agent uses WebFetch, WebSearch, and file tools to autonomously collect structured data
4. Data is automatically exported to JSON and CSV formats

Output locations:

• ~/.reverse-api/runs/collected/{folder_name}/ (permanent storage)
• ./collected/{folder_name}/ (local copy with readable name)

Output files:

• items.json - Collected data in JSON format
• items.csv - Collected data in CSV format
• README.md - Collection metadata and schema documentation

Model Configuration: Collector mode uses the collector_model setting (default: claude-sonnet-4-6). This can be configured in ~/.reverse-api/config.json.

Example workflow:

$ reverse-api-engineer
> Find 3 JS frameworks

# Agent autonomously searches and collects data
# Data saved to: ./collected/js_frameworks/

🔧 Configuration

Settings stored in ~/.reverse-api/config.json:

{
  "agent_provider": "auto",
  "claude_code_model": "claude-sonnet-4-6",
  "collector_model": "claude-sonnet-4-6",
  "opencode_model": "claude-sonnet-4-6",
  "opencode_provider": "anthropic",
  "output_dir": null,
  "output_language": "python",
  "real_time_sync": true,
  "sdk": "claude"
}

Model Selection

Choose from Claude 4.6 models for API generation:

• Sonnet 4.6 (default): Balanced performance and cost
• Opus 4.6: Maximum capability for complex APIs
• Haiku 4.5: Fastest and most economical

Change in /settings or via CLI:

reverse-api-engineer manual --model claude-sonnet-4-6

If you use Opencode, look at the models.

Agent Configuration

Configure AI agents for autonomous browser automation.

Agent Providers:

• auto (default): Playwright MCP 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.
• chrome-mcp: Drives your real Chrome browser via Chrome DevTools MCP. Useful when you need existing sessions, cookies, or auth. Requires Chrome 146+ and Node.js 20.19+; enable auto-connect at chrome://inspect/#remote-debugging.

The agent's reasoning model is the same as the SDK model — see Model Selection.

Change in /settings → "agent provider"

SDK Selection

• Claude (default): Direct integration with Anthropic's Claude API
• OpenCode: Uses OpenCode SDK (requires OpenCode running locally)

Change in /settings or edit config.json directly.

Output Language

Control the programming language of generated API clients:

• python (default): Generate Python API clients
• javascript: Generate JavaScript API clients
• typescript: Generate TypeScript API clients

Change in /settings → "Output Language" or edit config.json:

{
  "output_language": "typescript"
}

Real-time Sync

Enable or disable real-time file synchronization during engineering sessions:

• Enabled (default): Files are synced to disk as they're generated
• Disabled: Files are written only at the end of the session

When enabled, you can see files appear in real-time as the AI generates them. This is useful for monitoring progress and debugging.

Change in /settings → "Real-time Sync" or edit config.json:

{
  "real_time_sync": false
}

💻 CLI Commands

Use these slash commands while in the CLI:

• /settings - Configure model, agent, SDK, and output directory
• /history - View past runs with costs
• /messages <run_id> - View detailed message logs
• /help - Show all commands
• /exit - Quit

Scripted / Agent Usage

The CLI subcommands can be driven by another agent or a wrapper script. Pass --no-interactive (and/or --json) so they fail fast instead of opening questionary prompts, and pipe the structured output into jq.

When --json is set: stdout contains exactly one JSON document (the final result), Rich logs and progress are diverted to stderr, and the process exits with a stable code.

# Run an autonomous agent capture and get a single JSON result on stdout
reverse-api-engineer agent \
  --prompt "capture the public jobs api" \
  --url https://example.com/jobs \
  --json | jq

# List runs / inspect a run as JSON (empty history -> [])
reverse-api-engineer list --json
reverse-api-engineer show <run_id> --json

# Run a generated script non-interactively
#   --no-interactive : never open the script-picker / install confirm
#   --auto-install   : install missing deps on retry without asking
reverse-api-engineer run <run_id> --file api_client.py \
  --no-interactive --auto-install -- --org acme

agent --json output schema

Field	Type	Notes
schema_version	int	Currently 1. Bumped on breaking changes.
status	"ok" | "error"	Top-level result.
run_id	string | null	Stable id for follow-up show / engineer / run calls.
prompt	string	The prompt that was passed in.
url	string | null	Optional starting URL.
mode	string | null	Provider used ("auto" for Playwright MCP, "chrome-mcp").
har_path	string | null	Absolute path to the captured HAR (recording.har).
script_path	string | null	Absolute path to the generated client when reverse engineering ran.
usage	object	Token + cost usage from the engineer SDK ({input_tokens, output_tokens, total_cost}).
error	string | null	Human-readable error message when status == "error".

Exit codes

Code	Meaning
0	Success.
1	Runtime error (capture or engineering failed; details in error).
2	Misuse — required arg missing under --no-interactive / --json.

For run, the exit code is the underlying script's return code on success, or 1 if no script was found, or non-zero if --no-interactive would otherwise have to prompt.

💡 Examples

Example: Reverse Engineering a Job Board API

$ reverse-api-engineer
> fetch all apple jobs from their careers page

# Browser opens, you navigate and interact
# Close browser when done

# AI generates:
# - api_client.py (full API implementation)
# - README.md (documentation)
# - example_usage.py (usage examples)

# Scripts copied to: ./scripts/apple_jobs_api/

Generated api_client.py includes:

• Authentication handling
• Clean function interfaces
• Type hints and docstrings
• Error handling
• Production-ready code

🛠️ Development

Setup

git clone https://github.com/kalil0321/reverse-api-engineer.git
cd reverse-api-engineer
uv sync

Run

uv run reverse-api-engineer

Build

./scripts/clean_build.sh

🔐 Requirements

• Python 3.11+
• Claude Code / OpenCode (for reverse engineering)
• Playwright browsers installed
• API key for agent mode (see Agent Configuration)

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.