MobMock

MobMock is a complete Kotlin port of the core engine behind RayBytes/ChatMock. It allows Android applications to natively authenticate with a ChatGPT Plus/Pro account using OAuth (PKCE) and make direct API calls to the undocumented Responses API (the same endpoint used by OpenAI's Codex CLI) for powerful, on-device AI functionality without requiring a traditional API key.

By moving ChatMock's Python logic directly into Android, MobMock acts as an embedded, native Android LLM provider that powers apps like MobClaw.

Background & Acknowledgement

This code is based entirely on the pioneering work done in the ChatMock repository by GitHub user RayBytes (Game_Time).

ChatMock originally implemented the complex logic required to intercept the OpenAI sign-in flow, capture the chatgpt_account_id, and properly stream the undocumented text/event-stream format from chatgpt.com/backend-api/codex/responses. MobMock translates this Python logic layer-by-layer into Kotlin, using OkHttp and Kotlinx.Serialization.

⚠️ Note & Limits: Just like the original ChatMock, this tool relies on undocumented, experimental endpoints (OpenAI-Beta: responses=experimental). It requires an active, paid ChatGPT account to function. This project is not affiliated with OpenAI, and exists as an educational exercise. Use responsibly and at your own risk.

Features

• Native Android Authentication: Built-in OAuthManager to generate PKCE challenges and process the callback, enabling a WebView-based login flow.
• Auth Storage: Parses JWTs to extract the chatgpt_account_id and securely stores tokens, automatically taking care of token refresh.
• Message Conversion: Transforms standard OpenAI-compatible messages and tool definitions into the proprietary Responses API schema using MessageConverter.
• Reasoning Capabilities: Native support for GPT-5 reasoning options (low, medium, high, xhigh) and thinking summaries using the ReasoningManager.
• Streaming Parser: Advanced SSE parsing in SseTranslator to handle the complex, multi-part chat response chunks directly within Kotlin coroutines.

Usage inside an App

val mobMock = MobMock(context)

// 1. Kick off Login (Returns a URL to load in a WebView)
val session = mobMock.startLogin()

// ... WebView intercepts redirect to http://localhost:1455/auth/callback?code=XXXX
mobMock.completeLogin(code, session)

// 2. Start a streamed conversation
mobMock.chatCompletionStream(
    model = "gpt-5",
    messages = listOf(
        mapOf("role" to "user", "content" to "Hello!")
    )
).collect { sseLine ->
    // Emits standard OpenAI SSE chunks that UI can easily consume
}

License

This software aligns with the original project and is released under the MIT License. See LICENSE for details.