One-command setup for AI-assisted development
Integrates powerful tools for working with AI coding agents:
- 🧠 Empirica - Epistemic self-assessment (track what AI knows/learns)
- 📋 Beads - Git-backed issue tracker with dependency graphs
- 🎨 Perles - Terminal UI with kanban boards and BQL query language
- 🧵 Mem0 - Universal memory layer for AI agents
- 💬 gptme - Personal AI assistant in your terminal
- 📝 OpenCommit - AI-generated git commit messages
- 🤖 Aider - AI pair programming (advanced users)
- 🔍 CodeRabbit CLI - AI code review (pre-commit)
See agent-tools.yaml.template for configuration options and .analysis-workspace/ANALYSIS.md for detailed analysis of all tools.
Two installation approaches available:
-
Global Mode (Recommended) - Install once, use everywhere
- Tools installed to
~/.agent-tooling/ - Use across all your projects
- Unified CLI:
agent-toolscommand - Centralized configuration with per-project overrides
- Tools installed to
-
Per-Project Mode - Self-contained installation
- Tools and config in each project
- Complete isolation per repository
- Traditional approach
See docs/guides/INSTALLATION-MODES.md for detailed comparison and migration guide.
# One-line install
curl -fsSL https://raw.githubusercontent.com/jesposito/agent-tooling-setup/main/install-global.sh | bash
# Then in any project:
agent-tools init# One-line install
curl -fsSL https://raw.githubusercontent.com/jesposito/agent-tooling-setup/main/install.sh | bash# Clone or download this repository
git clone https://github.com/jesposito/agent-tooling-setup.git
cd agent-tooling-setup
# Make installer executable
chmod +x install.sh
# Run in your project directory
cd /path/to/your/project
/path/to/agent-tooling-setup/install.sh# Create agent-instructions.md template
./install.sh --with-agent-instructions
# Skip tool installation (config files only)
./install.sh --skip-install
# Show help
./install.sh --helpIf you have a Claude Code extension installed in VS Code, just ask:
Install the agent tooling setup from https://github.com/jesposito/agent-tooling-setup
Run the installer to set up Empirica, Beads, and Perles for this project.
Claude will handle everything automatically!
Supported Extensions:
- Anthropic Claude Code (official)
- Claude Code YOLO (community)
Both extensions are fully supported with optimized VS Code settings. See CLAUDE_CODE.md for full details and example prompts.
- bd (Beads CLI) - Task tracking
- perles - Terminal UI
- empirica - Epistemic tracking
- mem0 - Memory layer (requires vector DB setup)
- gptme - Terminal AI agent (with optional extras: browser, datascience, server)
- opencommit (oco) - AI commit messages (manual or git hook mode)
- aider - AI pair programming (advanced, auto-commits)
- coderabbit-cli - AI code review (external service)
All optional tools are DISABLED by default for safety. See:
- agent-tools.yaml.template - Configuration options
- docs/guides/AI-TOOLS-GUIDE.md - Complete usage guide
- docs/examples/INTEGRATION-WORKFLOW.md - Real-world workflow example
Global Mode:
~/.agent-tooling/- All tools and shared config~/.agent-tooling/config/.agent-tools.yaml- Global configuration- Per-project:
.agent-tools.local.yaml(overrides global config)
Per-Project Mode:
docs/DEVELOPMENT.md- Quick reference for agentsAGENTS.md- Workflow and "landing the plane" checklist.gitattributes- Git merge strategy for Beads.beads/- Beads database and config.agent-tools.yaml- Project-specific configurationagent-instructions.md- Comprehensive development guidelines (optional, use--with-agent-instructions)- Based on production-tested template
- Includes customization comments
- See docs/guides/CUSTOMIZATION.md for adaptation guide
# Create epistemic tracking session
empirica session-create --ai-id claude-code --output json
# Document what you know
empirica preflight-submit -
# See available work
bd ready# Claim a task
bd update <id> --status in_progress
# Work on code...
# Visualize your board
perles
# Create new tasks as needed
bd create "Task description" --priority P1# Complete tasks
bd close <id>
# Document learnings
empirica postflight-submit -
# Sync with git
bd sync
# Push everything (mandatory!)
git push- Git repository (must be run in a git project)
- Platform: macOS, Linux, or Windows (WSL recommended)
- See docs/guides/WINDOWS-NOTES.md for Windows compatibility
- Python: 3.10-3.12 recommended (some tools require modern Python)
- Optional: Go 1.16+ (for Beads/Perles from source)
- Optional: Node.js 18+ (for OpenCommit if using AI commit messages)
# Run comprehensive diagnostics
./agent-tools-doctor.sh
# Check specific components
./agent-tools-doctor.sh --check-python
./agent-tools-doctor.sh --check-tools
# Generate health report
./agent-tools-doctor.sh --report# Update all tools safely
./agent-tools-update.sh
# Update specific tool
./agent-tools-update.sh --tool beads
# Dry run (see what would update)
./agent-tools-update.sh --dry-runFeatures:
- Pre/post update health checks
- Automatic backups before updates
- Rollback capability if issues detected
- Keeps last 5 backups
- Beads: https://github.com/steveyegge/beads
- Perles: https://github.com/zjrosen/perles
- Empirica: https://github.com/Nubaeon/empirica
Without these tools:
- ❌ Agents lose context between sessions
- ❌ TODOs scattered in markdown files
- ❌ No dependency tracking
- ❌ Hard to visualize task status
With this setup:
- ✅ Persistent memory across sessions (Empirica)
- ✅ Structured task tracking (Beads)
- ✅ Dependency-aware workflow (Beads)
- ✅ Visual kanban boards (Perles)
- ✅ Git-based collaboration
- ✅ Powerful BQL query language
┌─────────────┬──────────────┬──────────────┬──────────────┐
│ Blocked │ Ready │ In Progress │ Closed │
├─────────────┼──────────────┼──────────────┼──────────────┤
│ [P0] Fix DB │ [P1] Add API │ [P1] Update │ [P2] Tests │
│ │ [P2] Refactor│ docs │ [P3] Cleanup │
└─────────────┴──────────────┴──────────────┴──────────────┘
# Critical open bugs
type = bug and priority = P0 and status = open
# Ready work
status = open and ready = true
# Recent updates
updated >= -24h order by priority
# Epic with all children
type = epic expand down depth *
Claude Code users: Both Anthropic Claude Code and Claude Code YOLO are fully supported. See CLAUDE_CODE.md for installation prompts and workflow examples.
This project includes VS Code settings (.vscode/settings.json) that automatically configure both extensions with the right context files.
Add this to your agent's context:
This project uses Empirica, Beads, and Perles for task management.
See docs/DEVELOPMENT.md for quick reference.
See AGENTS.md for workflow requirements.
- Run
perlesto see your task board - Run
bd readyto see what's available to work on - Review
AGENTS.mdfor session completion checklist - Use
bd createto add tasks for agents
This repository includes production-tested templates ready to customize:
agent-instructions.md- Comprehensive template with [CUSTOMIZE] markers- See docs/guides/CUSTOMIZATION.md for detailed adaptation guide
- Based on real-world production usage
- Includes examples for various project types
docs/DEVELOPMENT.md- Agent quick reference (auto-generated by installer).beads/config.yaml- Beads configuration.perles/config.yaml- Perles theme and board layout
New to this? Check out docs/guides/CUSTOMIZATION.md for examples of how to adapt the template for:
- Full-stack web apps
- Python data pipelines
- React Native mobile apps
- Your specific project needs
# Remove configuration files
rm -rf .beads docs/DEVELOPMENT.md AGENTS.md
# Remove from git
git rm -r .beads docs/DEVELOPMENT.md AGENTS.md .gitattributes
# Uninstall tools (optional)
brew uninstall bd perles # macOS
pip uninstall empiricaThis installer must be run inside a git repository:
cd your-project
git init # if not already a repo
./install.shAdd to your PATH:
# For Beads/Perles (Go binaries)
export PATH="$HOME/go/bin:$PATH"
# For Empirica (Python)
export PATH="$HOME/.local/bin:$PATH"Install from GitHub:
pip3 install --user git+https://github.com/Nubaeon/empirica.gitThis is a template setup. Customize for your workflow!
To create your own version:
- Fork this repository
- Modify
install.shand templates - Update the one-line install URL
- Share with your team
MIT - Use freely, modify as needed
- Beads by Steve Yegge
- Perles by Zack Rosen
- Empirica - Epistemic assessment framework