varun86

llm-agents.nix

Nix packages for AI coding agents and development tools. Automatically updated daily.

0
0
0
public
Forked
View on GitHub
llm-agents.nix

Nix packages for AI coding agents and development tools. Automatically updated daily.

Available Tools

AI Coding Agents

amp - CLI for Amp, an agentic coding tool in research preview from Sourcegraph
antigravity-cli - CLI for Google Antigravity, an agentic development platform
claude-code - Agentic coding tool that lives in your terminal, understands your codebase, and helps you code faster
claw-code - Claude Code rewrite CLI built from the official claw-code Rust workspace
code - Fork of codex. Orchestrate agents from OpenAI, Claude, Gemini or any provider.
codex - OpenAI Codex CLI - a coding agent that runs locally on your computer
copilot-cli - GitHub Copilot CLI brings the power of Copilot coding agent directly to your terminal.
crush - The glamourous AI coding agent for your favourite terminal
cursor-agent - Cursor Agent - CLI tool for Cursor AI code editor
droid - Factory AI's Droid - AI-powered development agent for your terminal
eca - Editor Code Assistant (ECA) - AI pair programming capabilities agnostic of editor
forgecode - AI-Enhanced Terminal Development Environment - A comprehensive coding agent that integrates AI capabilities with your development environment
gemini-cli - AI agent that brings the power of Gemini directly into your terminal
gitclaw - Universal git-native multimodal AI agent (formerly gitagent)
goose-cli - CLI for Goose - a local, extensible, open source AI agent that automates engineering tasks
grok - Grok Build, xAI's agentic coding tool
iflow-cli - AI coding agent for the terminal with free model access via the iFlow platform
jules - Jules, the asynchronous coding agent from Google, in the terminal
junie - Junie, JetBrains AI coding agent CLI
kilocode-cli - The open-source AI coding agent. Now available in your terminal.
letta-code - Memory-first coding agent that learns and evolves across sessions
mistral-vibe - Minimal CLI coding agent by Mistral AI - open-source command-line coding assistant powered by Devstral
nanocoder - A beautiful local-first coding agent running in your terminal - built by the community for the community ⚒
oh-my-codex - Multi-agent orchestration layer for OpenAI Codex CLI
oh-my-opencode - The Best AI Agent Harness - Multi-Model Orchestration for OpenCode
omp - A terminal-based coding agent with multi-model support
opencode - AI coding agent built for the terminal
openfang - Open-source Agent OS built in Rust — CLI for the OpenFang platform
pi - A terminal-based coding agent with multi-model support
qoder-cli - Qoder AI CLI tool - Terminal-based AI assistant for code development
qwen-code - Command-line AI workflow tool for Qwen3-Coder models
reasonix - DeepSeek-native AI coding agent for your terminal

AI Assistants

aionui - Desktop and WebUI cowork app that turns AI agents into a local assistant and server
hermes-agent - Self-improving AI agent by Nous Research — creates skills from experience and runs anywhere
hermes-desktop - Desktop companion for Hermes Agent
hermes-hud - TUI consciousness monitor for Hermes Agent
localgpt - Local AI assistant with persistent markdown memory, autonomous tasks, and semantic search
openclaw - Your own personal AI assistant. Any OS. Any Platform. The lobster way
picoclaw - Tiny, fast, and deployable anywhere — automate the mundane, unleash your creativity
vessel-browser - Agent-oriented browser with durable state and MCP control
zeroclaw - Fast, small, and fully autonomous AI assistant infrastructure

Claude Code Ecosystem

aperant - Autonomous multi-agent coding framework powered by Claude AI
catnip - Developer environment that's like catnip for agentic programming
cc-switch-cli - CLI version of CC Switch - All-in-One Assistant for Claude Code, Codex & Gemini CLI
ccstatusline - A highly customizable status line formatter for Claude Code CLI
claude-code-router - Use Claude Code without an Anthropics account and route it to another LLM provider
oh-my-claudecode - Multi-agent orchestration system for Claude Code

ACP Ecosystem

claude-agent-acp - An ACP-compatible coding agent powered by the Claude Code SDK (TypeScript)
codex-acp - An ACP-compatible coding agent powered by Codex

Usage Analytics

agentsview - Local-first viewer and analytics for AI coding agent sessions
ccusage - Analyze coding agent CLI token usage and costs from local data
entire - CLI tool that captures AI agent sessions and links them to code changes

Workflow & Project Management

agent-deck - Your AI agent command center
backlog-md - Backlog.md - A tool for managing project collaboration between humans and AI Agents in a git ecosystem
beads - A distributed issue tracker designed for AI-supervised coding workflows
beads-rust - Fast Rust port of beads - a local-first issue tracker for git repositories
beads-viewer - Graph-aware TUI for the Beads issue tracker
bernstein - Multi-agent orchestrator for CLI coding agents — spawn, coordinate, and manage parallel AI agents
but - GitButler CLI - virtual branches and AI-assisted Git workflow from the terminal
cc-sdd - Spec-driven development framework for AI coding agents
chainlink - Simple, lean issue tracker CLI designed for AI-assisted development
gastown - Gas Town - multi-agent workspace manager
gitbutler - Git client for simultaneous branches on top of your existing workflow
gnhf - Ralph/autoresearch-style orchestrator that keeps coding agents running while you sleep
herdr - Terminal workspace manager for AI coding agents
mardi-gras - Terminal UI for Beads issue tracking with a parade-inspired workflow view
openspec - Spec-driven development for AI coding assistants
openspecui - Visual interface for spec-driven development
paseo-desktop - Voice-controlled desktop development environment for AI coding agents
ralph-tui - AI Agent Loop Orchestrator TUI
sidecar - Terminal-based development companion for AI coding agents
spec-kit - Specify CLI, part of GitHub Spec Kit. A tool to bootstrap your projects for Spec-Driven Development (SDD)
td - A minimalist CLI for tracking tasks across AI coding sessions.
vibe-kanban - Kanban board to orchestrate AI coding agents like Claude Code, Codex, and Gemini CLI
workmux - Git worktrees + tmux windows for zero-friction parallel dev

Code Review

code-review-graph - Local knowledge graph for AI coding agents — builds persistent map of your codebase for token-efficient code reviews
coderabbit-cli - AI-powered code review CLI tool
cubic - AI code review CLI from cubic.dev - fast pre-flight review before you push
hunk - Terminal diff viewer for agentic changesets
tuicr - Review AI-generated diffs like a GitHub pull request, right from your terminal

Memory & Code Intelligence

ck - Local first semantic and hybrid BM25 grep / search tool for use by AI and humans!
codegraph - Semantic code intelligence for AI coding agents
context-hub - CLI for Context Hub - search and retrieve LLM-optimized docs and skills
gitnexus - Graph-powered code intelligence for AI agents
gno - Local-first knowledge engine with hybrid search, RAG Q&A, and MCP server integration
memvid-cli - AI memory CLI - crash-safe, single-file storage with semantic search
qmd - mini cli search engine for your docs, knowledge bases, meeting notes, whatever. Tracking current sota approaches while being all local
semble - Fast and accurate local code search for AI agents — CLI and MCP server
zat - Code outline viewer for LLM coding agents — shows exported symbols with line numbers

Sandboxing & Isolation

claudebox - Sandboxed environment for Claude Code
fence - Lightweight, container-free sandbox for running commands with network and filesystem restrictions
nono - Kernel-enforced agent sandbox. Capability-based isolation with secure key management, atomic rollback, cryptographic immutable audit chain of provenance. Run your agents in a zero-trust environment.
sandbox-runtime - Lightweight sandboxing tool for enforcing filesystem and network restrictions

Skills & Plugins

claude-plugins - CLI tool for managing Claude Code plugins
openskills - Universal skills loader for AI coding agents - install and load Anthropic SKILL.md format skills in any agent
skills - The open agent skills tool for installing and managing skills across AI coding agents
skills-installer - Install agent skills across multiple AI coding clients

Utilities

agent-browser - Headless browser automation CLI for AI agents
apm - Agent Package Manager — dependency manager for AI agents
cli-proxy-api - Unified proxy providing OpenAI/Gemini/Claude/Codex compatible APIs for AI coding CLI tools
codex-auth - CLI tool for switching Codex accounts
copilot-language-server - GitHub Copilot Language Server - AI pair programmer LSP
git-surgeon - Git primitives for autonomous coding agents
handy - Fast and accurate local transcription app using AI models
happy-coder - Mobile and Web client for Codex and Claude Code, with realtime voice and encryption
mcporter - TypeScript runtime and CLI for the Model Context Protocol
officecli - CLI for creating and editing Office Open XML documents
parallel-cli - AI-powered web search, extraction, and research CLI from Parallel
rtk - CLI proxy that reduces LLM token consumption by 60-90% on common dev commands
showboat - Create executable demo documents showing and proving an agent's work
toon - Rust implementation of TOON - Token-Oriented Object Notation for LLM prompts

Installation

Using Nix Flakes (Recommended)

Add to your system configuration:

{
  inputs = {
    llm-agents.url = "github:numtide/llm-agents.nix";
  };

  # In your system packages:
  environment.systemPackages = with inputs.llm-agents.packages.${pkgs.stdenv.hostPlatform.system}; [
    claude-code
    opencode
    gemini-cli
    qwen-code
    # ... other tools
  ];
}

[!NOTE]
This flake is only built and tested against its pinned nixpkgs-unstable
input. If you set llm-agents.inputs.nixpkgs.follows = "nixpkgs", your
nixpkgs must also track nixpkgs-unstable and be reasonably current —
using a stable release branch (e.g. nixos-25.05) will break eventually.
Omitting follows costs you a second nixpkgs evaluation but guarantees you
get the combination we ship in CI — and lets you pull pre-built binaries
from our binary cache instead of rebuilding everything
against your nixpkgs.

Using Overlay

Alternatively, use an overlay to access packages under the llm-agents
namespace. Two are provided:

  • overlays.default exposes packages.${system} as-is. Packages are built
    against this flake’s pinned nixpkgs, so the binary cache
    hits regardless of your nixpkgs revision, at the cost of evaluating a second
    nixpkgs instance.
  • overlays.shared-nixpkgs rebuilds each package against your nixpkgs, so
    dependencies are shared with the rest of your system and no extra nixpkgs is
    evaluated. The binary cache will only hit when your nixpkgs revision matches
    ours.
{
  inputs = {
    nixpkgs.url = "github:NixOS/nixpkgs/nixpkgs-unstable";
    llm-agents.url = "github:numtide/llm-agents.nix";
  };

  outputs = { nixpkgs, llm-agents, ... }: {
    # NixOS / nix-darwin configuration
    nixosConfigurations.myhost = nixpkgs.lib.nixosSystem {
      system = "x86_64-linux";
      modules = [{
        nixpkgs.overlays = [ llm-agents.overlays.default ];
        environment.systemPackages = [
          pkgs.llm-agents.claude-code
          pkgs.llm-agents.codex
          pkgs.llm-agents.gemini-cli
        ];
      }];
    };
  };
}

Try Without Installing

Browse all available tools with the interactive launcher:

nix run github:numtide/llm-agents.nix

This opens an fzf picker listing every package with its description.
Select one and it will be run via nix run.

Or run a specific tool directly:

nix run github:numtide/llm-agents.nix#claude-code
nix run github:numtide/llm-agents.nix#opencode
nix run github:numtide/llm-agents.nix#gemini-cli
nix run github:numtide/llm-agents.nix#qwen-code
# etc...

Binary Cache

Pre-built binaries are available from the Numtide binary cache. All packages are built daily via CI and pushed to the cache, so you can avoid compiling from source.

This cache is automatically configured when this flake is used directly (e.g nix run github:numtide/llm-agents.nix#claude-code)

To use the binary cache when using this flake as an input, add nixConfig to your flake:

{
  nixConfig = {
    extra-substituters = [ "https://cache.numtide.com" ];
    extra-trusted-public-keys = [ "niks3.numtide.com-1:DTx8wZduET09hRmMtKdQDxNNthLQETkc/yaX7M4qK0g=" ];
  };
}

Alternatively, you can configure this system-wide in your NixOS configuration:

nix.settings = {
  extra-substituters = [ "https://cache.numtide.com" ];
  extra-trusted-public-keys = [
    "niks3.numtide.com-1:DTx8wZduET09hRmMtKdQDxNNthLQETkc/yaX7M4qK0g="
  ];
};

Development

Setup Development Environment

nix develop

Building Packages

# Build a specific package
nix build .#claude-code
nix build .#opencode
nix build .#qwen-code
# etc...

Code Quality

# Format all code
nix fmt

# Run checks
nix flake check

Package Details

Platform Support

All packages support:

  • x86_64-linux
  • aarch64-linux
  • x86_64-darwin
  • aarch64-darwin

Experimental Features

This repository serves as a laboratory for exploring how Nix can enhance AI-powered development:

Current Experiments

  • Sandboxed execution: claudebox demonstrates transparent, sandboxed AI agent execution
  • Provider abstraction: claude-code-router explores decoupling AI interfaces from specific providers
  • Tool composition: Investigating how multiple AI agents can work together in Nix environments

Contributing

Contributions are welcome! Please:

  1. Fork the repository
  2. Create a feature branch
  3. Run nix fmt before committing
  4. Submit a pull request

See also

License

Individual tools are licensed under their respective licenses.

The Nix packaging code in this repository is licensed under MIT.

v0.3.3[beta]