# Camelid

A Rust-native local inference application for chat, file work, and coding with GGUF models on your own hardware, shipping as a single binary with a desktop app, browser UI, terminal, and OpenAI-compatible API, and no Python or Node.js at runtime.

Camelid packages a language-model inference engine and its web UI into one Rust binary, using Metal on Apple Silicon, CUDA on supported NVIDIA setups, or CPU fallback elsewhere, and validates supported model files and quantizations against a pinned llama.cpp reference with documented limits per configuration. Everyday chat supports streaming replies with markdown, code highlighting, math, and diagrams, editing and resending messages, and organizing conversations into projects with shared instructions and reference files.

Its code-in-chat feature works in a local project with a live plan, read-only helper agents, individually approved commands and checks, and a managed local preview server for Chrome, with the ability to pause, steer, queue a follow-up, or return to a saved session. Local or remote MCP servers can be connected with tool choice and approval per conversation, and a web-research mode does research with sourced replies.

It also offers a mobile-browser chat mode where inference stays on the desktop machine while the phone acts as a thin client, and runs entirely offline once model files are downloaded.

Website: https://github.com/timtoole02/Camelid