Mimic Uses Claude to Turn App Traffic Into Python Clients

A new MIT-licensed open-source project named Mimic has gathered over 2,600 stars on GitHub by streamlining how developers reverse-engineer private APIs. Instead of manually sifting through proxy logs, copying authorization headers, and hand-crafting request wrappers, Mimic intercepts an application's HTTPS network traffic and uses Claude to automatically construct a functional Python client.

The tool relies on a simple three-step workflow executed through the commands record, learn, and gen. For iOS applications, Mimic uses mitmproxy to capture network activity, while web applications can be ingested by pasting cURL commands or uploading HAR files. Claude then analyzes the recorded endpoints, request bodies, and responses, outputting standard Python code built on top of mimic.App. The generated client includes named methods, path parameters, request-body templates, and chained calls that pass data between dependent endpoints.

To maintain active connections, Mimic automatically extracts recurring session values like cookies, device identifiers, and authentication tokens. If an endpoint returns a 401 error, the client attempts to auto-refresh rotated tokens for idempotent requests, maintaining session persistence without manual intervention.

Despite its automation capabilities, Mimic faces technical boundary conditions. Developers encounter blocks from mobile certificate pinning, though this can be bypassed using Frida, as well as sender-constrained DPoP tokens, which currently have no workaround. Furthermore, the generated client is strictly bound to the specific endpoints triggered during the recording phase. Designed exclusively for personal accounts and data, Mimic turns tedious network inspection into an automated code-generation process.