FuXi
FuXi is a fast, self-contained terminal AI coding agent intended to turn ideas into working software by reading code, editing files, running commands, and orchestrating tools. Its stated vision is to let everyone with an idea become a creator, without requiring a technical background.
At a Glance
- Developers
- Professional software users and engineering teams
- Individuals, teams, and enterprises seeking a terminal coding agent
- Users of OpenAI-compatible and other LLM providers who want provider portability
AI Tools by FuXi
(1)
FuXi
Terminal AI Coding Agent in Go
Discussions
No discussions yet
Be the first to start a discussion about FuXi
Latest News
FuXi 0.1.8 released with rate-limit banners, official plugin marketplace mirror, long-session performance improvements, security fixes, Windows improvements, and MCP updates.
FuXi 0.1.6 released with unified session/message handling, hardened login, model validation, cross-platform improvements, and stability fixes.
FuXi announced and migrated its official plugin marketplace, including a move to the .fuxi-plugin manifest format.
FuXi 0.1.2 released with multi-window coordination, image captioning/vision detection, adaptive thinking, native voice/browser bridges, and plugin security hardening.
Products & Services
A proprietary, self-contained Go terminal/TUI agent distributed as a single static binary for macOS, Linux, and Windows. It reads and edits code, executes shell commands, uses MCP/browser/Jupyter/LSP tools, maintains sessions and memory, and routes requests across model providers with automatic failover. Users can sign in for FuXi-managed models or bring their own provider key.
An account-based managed-model path inside the FuXi service; pricing describes usage-based model calls and automatic matching of lightweight, standard, and high-performance capability tiers.
An official marketplace/mirror for installing FuXi plugins; the repository later migrated marketplace manifests and tooling to the .fuxi-plugin format.
Market Position
FuXi positions itself as a terminal-first, provider-agnostic, self-contained alternative to Claude Code, with cost-aware routing, automatic failover, local-first privacy, and a single static binary. Its own documentation compares it with Claude Code; broader comparison material also places it alongside OpenCode, OpenAI Codex, Gemini CLI, and other autonomous coding agents. Unlike a model-specific assistant, FuXi emphasizes bringing any compatible model and adding the agent loop, tools, memory, and safety controls around it.
Founding Story
FuXi was started as a terminal-first, provider-agnostic alternative to hosted coding agents: a single Go binary that puts an agentic Think–Act–Verify loop, tool orchestration, routing, persistence, and safety controls on the user's machine. The website frames the initial vision as removing the coding barrier so anyone can create; the public repository says the product source is proprietary and that its roughly two million lines of code were written by FuXi itself.
Business Model
Revenue Model
Usage-based billing for model calls. Users may bring their own provider key, while signed-in users can use FuXi-managed models; FuXi automatically maps tasks to capability tiers and charges by tokens. The software binary itself is described as free forever, with no license cost for individuals, teams, or enterprises.
Pricing Tiers
Free for everyone; terminal coding agent, code generation/editing/refactoring, file and command tools, session memory, MCP/plugins, pay only for usage, automatic model-tier matching, and usage dashboard.
$15 complimentary usage credit on signup, with more model capacity and multi-model usage-based calls.
Top-up pricing and higher usage capacity were listed as coming soon.
Higher limits/output capacity, early access to advanced capabilities, and priority access were listed as coming soon.
Target Markets
- Developers
- Professional software users and engineering teams
- Individuals, teams, and enterprises seeking a terminal coding agent
- Users of OpenAI-compatible and other LLM providers who want provider portability
- Natural-language software development in a terminal
- Code reading, generation, editing, refactoring, debugging, and test fixing
- Long-running multi-file or large-repository tasks
- Automated shell and developer-tool workflows with permission controls
- Cost-sensitive coding assistance using automatic model-tier selection
- Coding across provider outages using automatic failover