Agent Setup

Portable multi-agent configuration for Claude Code, Gemini CLI, OpenCode, and Codex. Includes skills, commands, and specialized agents for orchestrated AI workflows.

Prerequisites

Runtime Requirements

Requirement	Windows	macOS	Linux
Node.js 18+	`winget install OpenJS.NodeJS.LTS`	`brew install node`	`sudo apt install nodejs`
Python 3.11+	`winget install Python.Python.3.11`	`brew install python@3.11`	`sudo apt install python3.11`
Go 1.21+	`winget install GoLang.Go`	`brew install go`	`sudo apt install golang-go`
uv (Python package manager)	`pip install uv`	`brew install uv`	`pip install uv`

AI Agent Platforms

Claude Code (Anthropic)

Primary orchestrator for all commands and skills.

# All platforms
npm install -g @anthropic-ai/claude-code

# Verify
claude --version

Gemini CLI (Google)

Google AI models including Gemini 3 Pro and Computer Use.

# All platforms
npm install -g @anthropic-ai/gemini-cli

# Verify
gemini --version

Codex (OpenAI)

OpenAI models including GPT-4o and GPT-5.

# All platforms
npm install -g @openai/codex

# Verify
codex --version

OpenCode

Multi-model CLI for GLM, Grok, BigPickle, and MiniMax models.

# All platforms (requires Go)
go install github.com/opencode-ai/opencode@latest

# Verify
opencode --version

# Enable autonomous mode
export OPENCODE_YOLO=true

Available Models:

Model ID	Purpose
`opencode/big-pickle`	Deep analysis, edge cases
`opencode/grok-code`	Fast search, coherence checks
`opencode/minimax-m2.1`	Additional perspective

Cursor CLI

Cursor's agent mode via command line. Used by /hive-experimental and /swarm-experimental for Opus-powered agents.

# All platforms (requires Cursor installed)
# CLI is bundled with Cursor - add to PATH

# Windows (PowerShell)
$env:PATH += ";$env:LOCALAPPDATA\Programs\cursor\resources\app\bin"

# macOS
export PATH="$PATH:/Applications/Cursor.app/Contents/Resources/app/bin"

# Linux
export PATH="$PATH:/opt/cursor/resources/app/bin"

# Verify
cursor --version

Agent Mode:

# Run agent with prompt
cursor --agent "Implement feature X"

# With specific model
cursor --agent --model claude-opus-4 "Refactor this module"

Notes:

Requires Cursor app installed (https://cursor.sh)
Agent mode uses your Cursor subscription
Supports Claude, GPT-4, and other models configured in Cursor

Process Orchestration

mprocs

Required for /hive commands that run multiple agents in parallel.

Windows:

# Via Scoop (recommended)
Set-ExecutionPolicy RemoteSigned -Scope CurrentUser
irm get.scoop.sh | iex
scoop install mprocs

# Or via winget
winget install pvolok.mprocs

# Or via Cargo
cargo install mprocs

macOS:

# Via Homebrew
brew install mprocs

# Or via Cargo
cargo install mprocs

Linux:

# Via Cargo (requires Rust)
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
cargo install mprocs

# Or download binary
curl -fsSL https://github.com/pvolok/mprocs/releases/latest/download/mprocs-linux-amd64 -o /usr/local/bin/mprocs
chmod +x /usr/local/bin/mprocs

Verify:

mprocs --version

Development Tools

Tool	Purpose	Windows	macOS	Linux
gh	GitHub CLI	`winget install GitHub.cli`	`brew install gh`	`sudo apt install gh`
git	Version control	`winget install Git.Git`	`brew install git`	`sudo apt install git`
jq	JSON processing	`winget install jqlang.jq`	`brew install jq`	`sudo apt install jq`

Optional Dependencies

Text-to-Speech

Used by /tts and /speak commands.

ElevenLabs:

pip install elevenlabs
# Requires ELEVENLABS_API_KEY

Cartesia:

pip install cartesia
# Requires CARTESIA_API_KEY

Browser Automation

Used by web-app-testing skill and Gemini Computer Use.

Playwright:

pip install playwright
playwright install chromium

Web Scraping

Used by various skills for documentation fetching.

Firecrawl:

pip install firecrawl-py
# Requires FIRECRAWL_API_KEY

Environment Variables

Create a .env file or set these in your shell profile. See .env.example for a template.

Required

# Anthropic (Claude Code)
export ANTHROPIC_API_KEY="sk-ant-..."

# Google (Gemini CLI)
export GOOGLE_API_KEY="..."
export GEMINI_API_KEY="..."

# OpenAI (Codex)
export OPENAI_API_KEY="sk-..."

# OpenCode
export OPENCODE_API_KEY="..."
export OPENCODE_YOLO=true  # Enable autonomous mode

Optional

# Text-to-Speech
export ELEVENLABS_API_KEY="..."
export CARTESIA_API_KEY="..."

# Web Services
export FIRECRAWL_API_KEY="..."
export APIFY_API_KEY="..."  # LinkedIn scraping

Installation

Copy contents to your .claude/ directory:

macOS/Linux:

cp -r skills/* ~/.claude/skills/
cp -r commands/* ~/.claude/commands/
cp -r agents/* ~/.claude/agents/
cp CLAUDE.md ~/CLAUDE.md

Windows (PowerShell):

Copy-Item -Recurse skills\* $env:USERPROFILE\.claude\skills\ -Force
Copy-Item -Recurse commands\* $env:USERPROFILE\.claude\commands\ -Force
Copy-Item -Recurse agents\* $env:USERPROFILE\.claude\agents\ -Force
Copy-Item CLAUDE.md $env:USERPROFILE\CLAUDE.md -Force

Verification

After installation, verify all tools are working:

# AI Agent Platforms
claude --version
gemini --version
codex --version
opencode --version
cursor --version  # Optional - for /hive-experimental

# Orchestration
mprocs --version

# Development Tools
gh --version
git --version
jq --version

# Python Dependencies
python -c "import playwright; print('Playwright OK')"
python -c "import elevenlabs; print('ElevenLabs OK')"

Skill	Purpose
`agentic-patterns`	Design patterns for AI agents
`ai-sdk-best-practices`	Production guidelines for AI SDK
`curate-learnings`	Summarize learnings into project-dna
`dependency-verifier`	Validate package versions
`frontend-design`	Production-grade UI implementation
`gemini-frontend-design`	UI design with Gemini ideation
`gemini-image`	Image generation/analysis
`gemini-llm`	Gemini text generation
`gemini-video`	Video analysis
`openai-llm`	OpenAI text generation
`skill-creator`	Create new skills
`web-app-testing`	Browser automation with Gemini

Commands (44)

F-Thread (Fusion - Competing Implementations)

Command	Purpose
`/fusion-algorithm`	Compare algorithm implementations
`/fusion-api`	Compare API designs
`/fusion-arch`	Compare architecture patterns
`/fusion-bugfix`	Compare bug fix strategies
`/fusion-datamodel`	Compare data model designs
`/fusion-perf`	Compare performance optimizations
`/fusion-refactor`	Compare refactoring approaches
`/fusion-test`	Compare testing strategies
`/fusion-ui`	Compare UI designs

B-Thread (Hive - Multi-Agent Coordination)

Command	Purpose
`/hive`	Generic multi-agent (1-4 workers)
`/hive-experiment`	Spawn-on-demand experimental hive
`/hive-refactor`	9-agent large-scale refactoring
`/hive-dependabot`	Dynamic agents per Dependabot PR
`/create-hive-issue`	Create issue with hive
`/refine-hive-gitissue`	Refine issue with hive
`/resolve-hive-issue`	Resolve issue with hive (heavyweight)
`/resolve-hive-comments`	Resolve PR comments with hive (lightweight)
`/validate-hive-issue`	Validate issue with hive

C-Thread (Chained Workflows)

Command	Purpose
`/scout`	Multi-model codebase search
`/plan`	Multi-agent planning
`/build`	Feature implementation
`/scout_plan`	Scout -> Plan
`/scout_plan_build`	Scout -> Plan -> Build
`/scout_plan_build_test`	Scout -> Plan -> Build -> Test
`/scout_refine`	Scout -> Refine existing plan

L-Thread (Long-Running)

Command	Purpose
`/resolvedependabot`	Resolve all Dependabot PRs
`/resolveprcomments`	Resolve all PR comments
`/resolvegitissue`	Resolve GitHub issue

Compound Learning

Command	Purpose
`/fix`	Fix with learning injection
`/fix-hive`	Hive fix with compound learning
`/init-project-dna`	Bootstrap `.ai-docs/` structure

Utilities

Command	Purpose
`/assess-codebase`	Multi-agent codebase assessment
`/gitinfo`	Git repository info
`/create-issue`	Create GitHub issue
`/validateissue`	Validate issue structure
`/refine`	Refine existing plan
`/refinegitissue`	Refine GitHub issue
`/compare-contrast`	Compare options
`/test_features`	Browser testing
`/speak`, `/tts`	Text-to-speech

Agents (13)

Specialized Task agents for domain-specific work:

Agent	Domain
`ai-sdk-planner`	Vercel AI SDK architecture
`codebase-researcher`	Reverse-engineer codebases
`code-reviewer`	Code quality reviews
`design-review-agent`	UI/UX design reviews
`fastapi-specialist`	FastAPI development
`gcp-deployment`	Google Cloud deployment
`livekit-planner`	LiveKit real-time AI
`material3-expressive`	Material 3 design system
`mcp-server-specialist`	MCP server development
`pydantic-specialist`	Pydantic data modeling
`railway-specialist`	Railway deployment
`supabase-specialist`	Supabase development
`teams-integration-specialist`	Microsoft Teams integration

Hive Worker Models

Standard /hive and /resolve-hive-issue worker assignments:

Worker	Model	Role
worker-1	Claude Opus 4.5	Complex logic, architecture
worker-2	Gemini 3 Pro	Patterns, UI/frontend
worker-3	OpenCode Grok Code	Backend/frontend coherence
worker-4	Codex GPT-5.2	Code simplification
reviewer-bigpickle	OpenCode BigPickle	Edge cases, deep analysis
reviewer-grok	OpenCode Grok Code	Quick observations
resolver	Claude Opus 4.5	Address reviewer findings
tester	Codex GPT-5.2	Run tests, fix failures

Pre-Scan Agents (4 parallel)

Agent	Model	Focus
Agent 1	OpenCode BigPickle	Dependency analysis
Agent 2	OpenCode Grok Code	Code organization
Agent 3	OpenCode Grok Code	Quick search
Agent 4	OpenCode MiniMax M2.1	Config analysis

Compound Engineering

AI agents learn from past sessions to compound their effectiveness.

File Structure

Global (~/.ai-docs/):

universal-patterns.md - Cross-project patterns
model-insights.md - AI model capabilities
workflow-learnings.md - Thread type effectiveness
stopwords.txt - Keyword filtering

Per-Project (.ai-docs/):

learnings.jsonl - Append-only session learnings
project-dna.md - Curated project patterns
bug-patterns.md - Bug fix patterns

Learning Commands

Command	Pre-Session	Post-Session
`/fix`	Grep learnings	Append learning
`/fix-hive`	Grep learnings	Queen appends
`/hive`	Pre-scan greps	Queen appends
`/resolve-hive-issue`	Pre-scan greps	Queen appends

Bootstrap New Projects

/init-project-dna

Curate Learnings

/curate-learnings

Usage

# F-Thread: Compare approaches
/fusion-arch "e-commerce platform"

# B-Thread: Multi-agent work
/hive "feature-name" 3
/hive-refactor "src/legacy/"
/resolve-hive-issue 42

# C-Thread: Chained workflow
/scout_plan_build_test "Add user authentication"

# L-Thread: Long-running
/resolvedependabot

# Compound Learning: Fix with history
/fix "authentication bug in login flow"
/fix-hive "refactor auth module"

Platform-Specific Notes

Windows Path Handling

CRITICAL: mprocs on Windows requires Windows-style paths with escaped backslashes.

# CORRECT
cwd: "D:\\Code Projects\\MyProject"

# WRONG - Git Bash style will fail!
cwd: "/d/Code Projects/MyProject"

All hive commands use PowerShell to get proper Windows paths:

powershell -NoProfile -Command "(Get-Location).Path"

macOS ARM (Apple Silicon)

Some tools may need Rosetta or ARM-specific builds:

# If issues with Go tools
arch -arm64 go install github.com/opencode-ai/opencode@latest

Linux

Ensure ~/.local/bin and ~/go/bin are in your PATH:

export PATH="$HOME/.local/bin:$HOME/go/bin:$PATH"

Troubleshooting

Agent CLI not found

# Check if installed globally
npm list -g --depth=0

# Reinstall if missing
npm install -g @anthropic-ai/claude-code

mprocs fails on Windows

Use PowerShell, not Git Bash
Ensure paths use \\ not /
Run as Administrator if permission issues

OpenCode connection errors

# Verify API key
echo $OPENCODE_API_KEY

# Test connection
opencode run -m opencode/grok-code "Hello"

Python dependency issues

# Use uv for faster, reliable installs
uv pip install playwright elevenlabs cartesia

License

MIT

Name		Name	Last commit message	Last commit date
Latest commit History 3 Commits
.codex		.codex
.gemini		.gemini
agents		agents
commands		commands
global-learnings		global-learnings
hooks		hooks
output-styles		output-styles
skills		skills
status_lines		status_lines
swarm-templates		swarm-templates
.env.example		.env.example
.gitignore		.gitignore
CLAUDE.md		CLAUDE.md
README.md		README.md
settings.json		settings.json
settings.local.example.json		settings.local.example.json

Type	Pattern	Commands
Base	1 agent	Direct execution
P-Thread	Parallel	`/scout`, `/plan`
C-Thread	Chained	`/scout_plan_build_test`
F-Thread	Fusion	`/fusion-*`
B-Thread	Hive	`/hive`, `/hive-*`
L-Thread	Long-running	`/resolve*`

rdfitted/Claude-Code-Setup

Folders and files

Latest commit

History

Repository files navigation