Nyzhi is an AI coding agent built in Rust that connects to every major LLM. Give it a task, and it reads your code, writes changes, runs tests, and verifies results — all from the terminal.
What Nyzhi Does
Describe what you want in plain language. Nyzhi will:
- Read your codebase — files, diffs, search results, LSP diagnostics
- Write and edit files, apply patches, and refactor code
- Run shell commands, tests, builds, and linters
- Verify outcomes with structured evidence (build logs, test output, lint results)
It works in both interactive mode (a rich TUI) and non-interactive mode (nyz run), so you can chat in real time or run one-shot tasks from scripts and CI.
Why Nyzhi
Single binary, zero runtime deps. Built entirely in Rust, Nyzhi ships as a standalone executable. No Node, Python, or Docker required. Install once and it runs anywhere.
LLM-agnostic. Connect OpenAI, Anthropic, Gemini, OpenRouter, DeepSeek, Groq, and any OpenAI-compatible endpoint. Switch providers with a config change — or use routing to auto-select models by task complexity.
50+ built-in tools. File ops, git, grep, semantic search, LSP integration, web fetch, browser automation, PR creation, sub-agent delegation, team workflows, and more. See Tools for the full catalog.
MCP native. Model Context Protocol servers plug in seamlessly. Use the filesystem MCP, connect to remote APIs, or run your own — all without changing Nyzhi itself.
Autonomous execution. Autopilot runs tasks end-to-end: expansion, planning, execution, QA, and validation. Teams coordinate multiple agents with task boards, messaging, and file locking.
Rich TUI. Syntax highlighting, tab completion, session export, and theming. See TUI for details.
Core Concepts
- Sessions — Conversations persist in JSON. Save, load, search, and resume. See Sessions.
- Skills — Reusable prompt templates and workflows. See Skills.
- Memory — Project-scoped and user-scoped persistent memory. See Memory.
- Hooks — Run commands after edits, after turns, or around tool execution. See Hooks.
- Verification — Build/test/lint checks with structured evidence. See Verification.
Next Steps
- Getting Started — Install, connect a provider, and run your first task in under two minutes
- Configuration — Global, project, and local TOML config
- Architecture — How Nyzhi is built: six crates, one binary