Our Bold Vision ·
Let Everyone With IdeasBecome a True Creator
Some people's minds are full of creativity
But blocked by a wall called 'code'
Who has the right to create?
From today, the answer is only one — everyone
No coding required
No technical background needed
All you need is an idea
Leave the rest to FUXI
For everyone, no exceptions
Because the right to create belongs to all
FUXI · Install
Get running in under 30 seconds
macOS / Linux, Windows (PowerShell), Windows (CMD), no runtime required.
Quick Install FUXI
macOS / Linux, Windows (PowerShell), Windows (CMD)
curl -fsSL https://downloads.fuxicode.com/bootstrap.sh | bash
irm https://downloads.fuxicode.com/bootstrap.ps1 | iexcurl -fsSL https://downloads.fuxicode.com/install.cmd -o %TEMP%\install.cmd && %TEMP%\install.cmdFuXi Configuration File & Terminal Commands Guide
On first run, FuXi creates its configuration under ~/.fuxi/. You will need a chat-capable model to proceed.
providers: ds: type: openapi base_url: https://api.deepseek.com/v1 api_key: sk-xxx models: - id: deepseek-v4-pro-260425 model_canonical_name: deepseek-v4-pro ollama: type: openapi base_url: http://127.0.0.1:11434/v1 api_key: dummy models: - id: qwen14b model: active: { provider: ds, id: deepseek-v4-pro-260425 }
Configure model, then use /model to switch, /config for others – all via TUI slash commands (permissions, hooks, skills, plugins).
For details, please refer to the documentation
Multi-LLM Routing — One agent, all major models
ERNIE
Zhipu
DeepSeek
OpenAI
Kimi
Qwen
Doubao
Hunyuan
MiniMax· Domestic Models × FUXI
Give domestic LLMs Claude-level performance and capability
The model is just the engine — FUXI is the whole car. FUXI is a complete AI coding agent framework — equipping DeepSeek, GLM, Kimi, Qwen and other domestic LLMs with deep thinking, tool orchestration, sub-agents and memory management, delivering world-class coding performance. Data stays under your control, at a fraction of the cost.
Deep reasoning engine
Refactors and root-cause hunts — as deep as Claude
Deep Tool Suite
From chat completion to autonomous execution
Sub-Agent Orchestration
Even large repos, tackled methodically
Context & Memory
Multi-hour tasks, finished in one go
Permissions & Safety Rails
Let it run freely, without losing control
Full OpenAPI · Self-Controlled
World-class capability — data and cost in your own hands
Core Capabilities
One agent, eight capabilities
Everything you need, shipped as one binary — no extra dependencies.
Multi-LLM routing
One agent across Anthropic, OpenAI, Gemini, Bedrock, Vertex, GLM, DeepSeek, Grok, Qwen, Kimi.
Zero vendor lock-in — switch to the best model anytime
Intelligence layer
Complexity-based tiered routing, automatic failover, primary-backup racing.
Never stalls on rate-limits or outages — saves cost too
Deep tool suite
50+ tools: file edits, a shell safety classifier, code search, LSP diagnostics, browser.
The model acts in your real environment, not a sandbox
MCP client
Connect any MCP server (stdio / HTTP / WebSocket), OAuth supported.
The whole MCP ecosystem, plug-and-play
Extensible
Hooks, skills, plugins, custom slash commands — all hot-reloaded.
Tailor it to your workflow — no source changes needed
Persistent sessions & memory
JSONL transcripts, resume / rewind, cross-session dreaming memory, auto context compaction.
Long tasks never lose context — resume anytime
BYO key or sign in
Use a provider API key, or the FuXi OAuth hosted path.
Start with no key of your own
Multi-agent teams
Parallel sub-agents / swarm on a shared task list, leader-worker permissions.
Parallelize the work — big tasks finish fasterFUXI · Deep Comparison
FUXI vs. Real Products: Feature-by-Feature Comparison
Based on publicly available official documentation as of early 2026. 11 capabilities, 7 products.
Yes
Partial
Not supportedRustNodeElectronNodePythonREPLNot native
· Effect Analysis
Among these seven products, FUXI is the only one that automatically fails over to healthy nodes on provider outages and can race primary vs. backup models for the fastest response — the other six have no documented provider-level failover or racing. Combined with cost-aware tiered routing (only Cursor has partial implementation), FUXI leads uniquely in resilience.
No other product combines all three: multi-provider access, a single static binary with no runtime, and access for individuals, teams, and enterprises (including a hosted path requiring no API keys). Open-source alternatives (Codex CLI, Gemini CLI, Aider) either need a Python runtime or lack a hosted path.
On MCP, rich TUI, extensibility (hooks/skills/plugins), durable sessions, and shell safety, mainstream products have largely converged — FUXI matches the strongest, not stands alone. Competitors also have real strengths worth acknowledging: Codex CLI's open-source ecosystem, Cursor's IDE integration depth.
Source: early 2026 official docs. Capabilities subject to change.
A New Era
Starts Today
Download FUXI and run your AI coding agent in the terminal — no runtime, no complex setup, start coding instantly.
Get $15 usage on signup and start now
fuxi