Reverse API Engineer
Open-source CLI that captures browser traffic as HAR files and uses an AI model to generate typed API clients for websites.
At a Glance
MIT-licensed CLI that runs locally; no fees. Model/API usage for agent mode is an external cost.
Engagement
Available On
Alternatives
Listed Oct 2026
About Reverse API Engineer
Reverse API Engineer (rae) is an MIT-licensed command-line tool by kalil0321 that turns websites into API clients. You give it a site and a goal, a browser is driven by you or by an AI agent, traffic is captured to a HAR file, and your configured model writes a working client. The latest release listed on GitHub is v0.14.0, which added PowerShell support and Windows compatibility.
What It Is
Reverse API Engineer is a reverse-engineering assistant for undocumented HTTP APIs. Instead of opening DevTools, copying cURL commands and assembling a client by hand, you browse a site and the tool maps the backend calls the site actually makes. Output is plain source files written into your project, with a generated README and a runnable example, in Python, JavaScript, TypeScript, Go, Java, C#, PHP, Ruby, C, or PowerShell.
How the Capture Workflow Works
The CLI has four modes you cycle with Shift+Tab: manual (you drive a local Playwright browser), agent (an AI agent browses autonomously), engineer (re-run generation on a previous capture), and collector (an agent gathers structured JSON/CSV data using web search and fetch). Agent mode can use Playwright MCP, Chrome DevTools MCP (drives your real Chrome with existing sessions), or the Vercel agent-browser CLI.
Generated Output
The docs describe generated clients with type hints, authentication handling (cookies, bearers, CSRF tokens), pagination helpers, error types, and inline documentation. Runs are saved under ~/.reverse-api/runs/ with message logs, and token usage and cost are tracked per run.
Model and Agent Backends
The tool uses Claude by default and can route through OpenCode (including Ollama models via OpenCode), GitHub Copilot, or Cursor SDK backends when optional dependencies and credentials are configured. Scriptable subcommands (agent, engineer, run, list, show) support JSON output and NDJSON streaming so other agents or CI jobs can drive captures.
Tradeoffs to Know
The README advises reviewing generated code before executing it, and notes that sites with aggressive bot detection may block capture or need manual interaction. Intel macOS and 32-bit Windows are not supported installation platforms.
Community Discussions
Be the first to start a conversation about Reverse API Engineer
Share your experience with Reverse API Engineer, ask questions, or help others learn from your insights.
Pricing
Open Source
MIT-licensed CLI that runs locally; no fees. Model/API usage for agent mode is an external cost.
- MIT licensed
- Runs locally, no account, no server, no analytics
- Generates typed clients in Python, JavaScript, TypeScript, Go, Java, C#, PHP, Ruby, C, or PowerShell
- Manual, agent, engineer and collector modes
- Requires an API key for agent mode (external cost)
Capabilities
Key Features
- Captures browser traffic to HAR files
- Manual, agent, engineer and collector modes
- Playwright MCP, Chrome DevTools MCP and agent-browser providers
- Generates typed API clients in ten languages
- Handles auth, pagination and error types in generated code
- JSON and NDJSON output for scripting and agent workflows
- Session history and per-run token cost tracking
- Stealth patches for manual Playwright browsing
- Runs locally with MIT license
