Happy slice of mcp pizzaMCP.pizza

Fuxi

MCP.Pizza Chef: fuxicodex

FuXi is a coding assistant you run in a terminal window. It reads the files in a project, edits them, runs commands, and searches the web, asking your permission before it takes action. Any Model Context Protocol server you already use can be plugged in, and its tools become available to the assistant. Sign in with a FuXi account, or supply your own key from Anthropic, OpenAI, Google, or another provider. Conversations are saved, so you can pick a session back up later.

Coding

Use This MCP client To

Ask questions about an unfamiliar codebase from my terminal Have the assistant edit files and run commands for me Connect the add-ons I already use to one terminal assistant Switch between Anthropic, OpenAI, and Google models mid-task Resume yesterday's session and keep working where I left off Review each change before it touches my files

README

FuXi

English | 简体中文

An AI coding agent that lives in your terminal. Codename YiHuaKaiTian — "one stroke opens the heavens."

FuXi is a fast, self-contained AI developer terminal — a coding agent you drive from a rich TUI. It reads your codebase, edits files, runs commands, drives tools, and routes every request across multiple LLM providers with cost-aware routing and automatic failover. Written in Go, it ships as a single static binary with no runtime dependencies.

Homepage: https://www.fuxicode.com

Highlights

  • Multi-LLM routing — one agent, many providers. First-class support for:
    • Anthropic (native Messages API) and Anthropic on AWS Bedrock / Google Vertex
    • OpenAI-compatible endpoints (OpenAI, Azure OpenAI, and any OpenAI-style API)
    • Google Gemini
    • Popular open/China-hosted models via OpenAI-compatible endpoints — **GLM **, DeepSeek, Qwen, Hunyuan, Doubao, Ernie, Grok (xAI), and more
  • Intelligence layer — cost-aware routing, automatic failover to a healthy provider, optional racing of primary vs. fallback, and role-based model tiers (fast / primary / reasoning) resolved from your config.
  • A deep tool suite — file read/edit/write, shell (bash / PowerShell) with a safety classifier, web fetch, code search, LSP-backed diagnostics, Jupyter notebooks, computer/browser use via MCP, background tasks, and parallel sub-agents.
  • MCP client — connect any Model Context Protocol server (stdio, HTTP, or WebSocket) and its tools become available to the agent.
  • Extensible — hooks, skills, plugins, and user-defined slash commands.
  • Durable sessions — persistent transcripts, checkpoints/resume, an idle "dreaming" memory, and automatic context compaction for long conversations.
  • Bring your own key, or log in — use a provider API key, or sign in with FuXi OAuth. Keys are never required for the FuXi-managed path.
  • Self-updating — a background version check and a one-command fuxi update keep your install current, with checksum verification before it ever replaces the running binary.

Install

macOS / Linux

curl -fsSL https://releases.fuxicode.com/bootstrap.sh | bash

Windows (PowerShell)

Fuxi FAQ

Which apps does this work in?
It is its own terminal program — you run FuXi directly rather than adding it to another assistant.
Do I need my own key?
Not necessarily. You can sign in with a FuXi account, or bring a key from a provider you already pay for.
Can I use this to connect the add-ons I already have?
Yes — it loads any Model Context Protocol server and makes its tools available while you work.
How hard is the setup?
You run one install command in a terminal, then either sign in or paste in a provider key.
Which computers does it run on?
macOS, Linux, and Windows. It ships as a single program with nothing else to install alongside it.
Can it use more than one model provider?
Yes — it supports Anthropic, OpenAI-compatible services, and Google Gemini, and can fall back to a healthy one automatically.