kalil0321/reverse-api-engineer

Reverse API Engineer – Turn a website into a ready‑to‑use API client

What it does

• You point the tool at a web page and tell it what you want (e.g. fetch all Apple jobs).
• It opens a browser, either you control it manually or an AI‑driven “agent” does the navigation.
• While the page loads it records all network requests into a HAR (HTTP Archive) file.
• A language model (Claude, OpenCode, Ollama, Copilot, etc.) reads that traffic and writes a typed client library in the language you choose (Python, JavaScript/TypeScript, Go, Java, C#, PHP, Ruby, or even C).
• The generated client is saved in a folder with a small README and an example script, ready to be imported into your own code.

Why it matters

• Avoid the tedious “open DevTools → copy cURL → hand‑craft request” loop.
• Works for sites that expose undocumented JSON endpoints, letting you treat them like a proper API.
• Supports both interactive capture (you click around) and fully automated capture via an AI agent that can log in, click, scroll, etc.

Installation

# Using uv (recommended) or pip
uv tool install reverse-api-engineer   # or pip install reverse-api-engineer

• Agent mode (default) needs only the core package.
• Manual mode requires the optional [manual] extra and a Playwright Chromium install:

uv tool install "reverse-api-engineer[manual]"
playwright install chromium

Quick start

reverse-api-engineer
> fetch all apple jobs from their careers page
# a browser window opens – navigate, interact, then close
# generated code appears in ./scripts/apple_jobs_api/

You can switch modes on‑the‑fly with Shift+Tab.

Modes

Mode	How it works
manual	You drive the browser; the AI only generates the client after capture.
agent	An AI agent drives the browser automatically (Playwright MCP, Chrome‑MCP, or Vercel’s agent-browser).
engineer	Re‑run generation on a previously saved HAR capture.
collector	The agent also extracts structured data (JSON/CSV) using web search + fetch.

Configuration

• Settings live in ~/.reverse-api/config.json and are editable via the /settings command inside the CLI.
• You can pick the LLM provider (claude, opencode, cursor, copilot) and model (e.g. claude‑sonnet‑4‑6).
• Output language is selectable (python, javascript, typescript, go, java, csharp, php, ruby, c).
• The tool can start local Ollama or OpenCode servers automatically, or connect to existing ones.

CLI basics

• /settings – edit model, SDK, agent provider, output language, etc.
• /history – list past runs with timestamps, token usage, and cost.
• /messages <run_id> – view the full conversation with the model for a run.
• /help – show all commands.
• Non‑interactive / scriptable usage via --json and --no-interactive flags, with JSON output schemas for easy piping to jq.

Output

• Permanent run data: ~/.reverse-api/runs/scripts/<run_id>/
• Friendly copy in your project: ./scripts/<descriptive_name>/
• Collector results: ./collected/<folder>/ containing items.json, items.csv, and a README.

Caveats

• The generated code is produced by a language model; always review it before execution.
• Sites that aggressively block bots may require manual interaction or additional workarounds.

Development

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

• Build script: ./scripts/clean_build.sh.
• Requires Python 3.11+, an API key for the chosen LLM, and (for manual mode) Playwright Chromium.

License – MIT.

---

Bottom line: Reverse API Engineer is a Python‑based CLI that captures a website’s network traffic, feeds it to a large language model, and spits out a ready‑to‑use typed API client in many popular programming languages. It bridges the gap between web scraping and proper API consumption, leveraging modern LLMs to automate the reverse‑engineering step.