ChatSpeed
A Rust-based desktop AI assistant and proxy that runs a tool-driven plan, execute, and review workflow across multiple language models, with a centralized MCP hub, sub-agent delegation, and persistent memory and skills.
ChatSpeed is a desktop AI assistant built in Rust around a tool-driven Plan, Execute, Final Review loop meant to keep long, multi-step coding and research tasks from stalling in chat. It splits work between a main agent and isolated child agents for parallel subtasks and review passes, and combines persistent memory, reusable skills, and MCP-based tool expansion into one execution system.
It supports OpenAI-compatible, Gemini, Claude, and Ollama models for planning, execution, utility, and vision roles, and a companion component adds tool-calling support to models that lack native tool calling. Beyond coding work, it doubles as a general desktop assistant for translation, mind maps, and flowcharts, and its MCP proxy exposes its own web search and fetch tools plus any installed MCP tools to external clients over Streamable HTTP.