abhishekk130804/claude-mythos-ai-anthropic-app

Claude Mythos APP presents itself as an unofficial desktop and mobile client for talking to Anthropic's Claude models. It is not made by Anthropic, and the README is upfront about that under a disclaimer. The app does not come with any API key. The user has to supply their own Anthropic API key, and all calls are billed to that key, which means you need an account with Anthropic to use it. The pitched use case is creative writing and roleplay rather than general chat. The README lists support for character cards, long stories, and immersive roleplay sessions. There are also features aimed at power users: management of custom system prompts and what the README calls jailbreak styles, reverse proxy support, local chat storage, image uploads, and document analysis. The interface is described as a dark Mythos aesthetic, with no screenshots shown in the README beyond a banner image. The stated tech stack mixes Python 3.12 and Rust on the backend with Next.js plus Electron or Tauri on the frontend, and integrates with the Anthropic API through LangChain. Installation for regular users is a ZIP download from the GitHub releases page, while developers are told to clone the repo with submodules, run pip install for Python dependencies, and start the dev server with npm run dev or python main.py. The README also includes recommended sampling settings: a temperature of 0.7 to 0.85 for creative work, 0.2 to 0.4 for coding, top_p of 0.95, and a max tokens range from 8192 up to 200000 depending on the model. A few things in the README, including the heavy keyword list in the GitHub description and a stray Russian sign-off at the bottom, suggest the project page is partly written for search engine reach. Readers should treat unofficial Claude clients with care: API keys and chat content pass through whatever code the app runs.