NYZHI v1.0.0

Built for >

The autonomous AI coding agent that lives in your terminal. Describe what you need, and watch Nyzhi read, write, test, and ship.

$ curl -fsSL get.nyzhi.com | sh

Also available via: cargo npm brew

nyzhi — bash

❯ nyz "Add authentication using JWT and Redis"

Analyzing project architecture... found Express.js

Writing src/middleware/auth.ts

edit +42 lines

Updating package.json

bash npm i jsonwebtoken redis

Task completed. Created commit "feat: add JWT auth via Redis"

❯

Read the Docs GitHub

What makes nyzhi different

One binary. Every LLM. Full autonomy.

Every model, one binary

OpenAI, Anthropic, Gemini, OpenRouter, DeepSeek, Groq — and any OpenAI-compatible endpoint. Switch mid-session. Route prompts to the right cost tier automatically.

Autonomous execution

Autopilot runs a five-phase pipeline: expand, plan, execute, QA, validate. Teams spawn coordinated sub-agents with mailbox messaging.

A TUI worth living in

8 themes. 14 accent colors. Syntax highlighting. Persistent history with reverse search. Desktop notifications.

MCP native

Drop a .mcp.json in your project root. Same format as Claude Code and Codex. stdio and HTTP transports. Deferred loading for fast startup.

Your code stays yours

Every file change is tracked. /undo reverts anything. Trust modes gate what runs without approval. The agent learns your patterns via /learn.

Every major LLM provider

Open-source first. Multi-account per provider. Automatic token refresh. Rate-limit rotation across accounts.

Provider	Models	Auth	Type
Kimi (Moonshot)	Kimi K2.5, K2, K2 Turbo	API key	Open
MiniMax	MiniMax-M2.5, M2.5 Highspeed, M2.1	API key	Open
GLM (Z.ai)	GLM-5, GLM-5 Code, GLM-4.7, GLM-4.7 FlashX	API key	Open
DeepSeek	DeepSeek V3.2 Chat, R1 Reasoner	API key	Open
Groq	Llama 3.3 70B, Llama 3.1 8B	API key	Open
Together AI	Llama 4 Maverick, Qwen3 235B, DeepSeek R1	API key	Open
Ollama	Qwen3 32B, Llama 3.3 70B, Devstral 24B	Local	Local
OpenAI	GPT-5.3 Codex, GPT-5.2, o3, o4-mini	API key / OAuth	Proprietary
Anthropic	Claude Opus 4.6, Sonnet 4.6, Haiku 4.5	API key / OAuth	Proprietary
Gemini	Gemini 3.1 Pro, 3 Flash, 2.5 Flash	API key / OAuth	Proprietary
Antigravity	Gemini 3.1 Pro, Claude Sonnet 4.6, GPT OSS 120B	API key (free tier)	Cloud
OpenRouter	Any model on the platform	API key	Aggregator
Custom	Any OpenAI-compatible endpoint	Configurable	Custom

13 built-in providers · bring your own OpenAI-compatible endpoint

50+ built-in tools

Zero runtime dependencies. A massive standard library of agent capabilities, all gated by a permission model you control.

Files

read write edit multi_edit apply_patch glob grep list_dir directory_tree file_info delete_file move_file copy_file create_dir

Shell

live streaming output

bash

Git

git_status git_diff git_log git_show git_branch git_commit git_checkout

Agent

task todo_write todo_read notepad_write notepad_read update_plan think load_skill tool_search

Analysis

verify lsp_diagnostics ast_search lsp_goto_definition lsp_find_references lsp_hover

Web & Browser

web_fetch web_search browser_open browser_screenshot browser_evaluate

Teams

team_create team_delete send_message task_create task_update task_list team_list read_inbox

PR & Search

create_pr review_pr semantic_search fuzzy_find instrument remove_instrumentation tail_file batch_apply

Six crates, zero cycles, one binary

Clean dependency graph. Built entirely in Rust for performance and safety.

Entry Point

nyzhi The CLI

Binary entry point. CLI parsing, command dispatch, tool and MCP assembly.

nyzhi-tui

Terminal UI. ratatui, themes, syntax highlighting, completion, history.

nyzhi-core

Agent loop, 50+ tools, sessions, workspace, MCP, planning, teams, hooks.

nyzhi-provider

LLM abstraction. OpenAI, Anthropic, Gemini with streaming and thinking.

nyzhi-auth

OAuth2 PKCE + device code, API keys, token store, multi-account rotation.

nyzhi-config

Config loading and merging across global, project, and local layers.

Command Line

Everything from a single command.

No bloated interfaces. Just nyz. Drop it into any directory, pass a prompt, and watch it work. Fully scriptable, fully composable.

nyz --help

nyz login <provider> OAuth login

nyz config Show merged config

nyz init Create .nyzhi/ in project

nyz mcp add|list|rm Manage MCP servers

nyz sessions [query] Search sessions

nyz cost [daily|weekly] Cost report

nyz deepinit Generate AGENTS.md

nyz update Self-update or rollback

Terminal

nyz interactive TUI

nyz run "<prompt>" one-shot (non-interactive)

nyz run -i img.png "<prompt>" with image

nyz -c / --continue resume last session

nyz -s / --session "<query>" resume by title or ID

nyz -p openai -m gpt-5.2 provider and model flags