Conduct CLI

AI Agent Orchestration Platform with Persistent Memory

Conduct v0.1 is a CLI for managing AI agents through spec-run-check workflows with a persistent libSQL memory database. Works with Cursor, Claude, Warp, Gemini, Factory, OpenCode, and GitHub Copilot.

Quick Start

# Install globally
npm install -g conduct-cli

# Or use install script
curl -fsSL https://git.conduct.run/install.sh | bash

# Initialize in your project
conduct init

What It Does

Conduct provides:

• Persistent Memory - libSQL database (local or remote) remembers all your work
• 11 CLI Commands - Complete memory management and orchestration
• 7+ AI Agents - Works with Cursor, Claude, Warp, Gemini, Factory, OpenCode, Copilot
• 9 Agent Templates - Structured commands for spec-run-check workflows
• Issue Tracker Integration - GitHub and Linear support with conduct config tracker
• JSON Schema Support - IDE autocomplete for config files
• Spec → Run → Check Workflow - Three-phase structured development

CLI Commands

Core Commands

conduct init        # Initialize project with memory database
conduct upgrade     # Update templates and instructions
conduct save        # Execute SQL operations (secure)
conduct health      # System health check
conduct config      # Configuration & tracker management

Integration

conduct sync        # Sync with GitHub/Linear issues
conduct config tracker add    # Configure issue trackers

Discovery & Maintenance

conduct discover    # Auto-discover features in codebase
conduct reconcile   # Detect drift between memory and code
conduct relevancy   # Calculate relevancy scores (time decay)
conduct archive     # Archive old runs
conduct list        # Query memory database

Memory Database

All work stored in libSQL (SQLite-compatible):

• Local Mode: _conduct/memory.db (solo development)
• Remote Mode: Turso/libSQL server (team collaboration)
• What's Stored: Features, specs, runs, checks, issue connections

Example:

conduct init
# Choose: Local (embedded) or Remote (Turso)

conduct list --specs
# 📋 Specs (3):
#   spec-1 [completed] Authentication System

conduct list --features
# ✨ Features (15):
#   #1 authentication - Auth System [active]

Agent Templates

Conduct installs 9 templates to guide AI agents:

1. conduct-spec - Create specification
2. conduct-run - Execute specification
3. conduct-check - Verify implementation
4. conduct-design - Create UI designs (optional)
5. conduct-index - Discover features
6. conduct-reconcile - Sync memory with codebase
7. conduct - Lightweight mode (no ceremony)
8. conduct-dry-run - Plan without executing
9. conduct-dry-check - Verify plan

Directory Structure

_conduct/
├── specs/          # Specifications
├── runs/           # Execution logs
├── checks/         # Verification reports
├── designs/        # UI designs
├── operations/     # SQL files for memory
├── templates/      # 9 agent command templates
└── memory.db       # libSQL database

conduct.json        # Configuration ($schema supported)
conduct.track.json  # Fast local tracker ($schema supported)
AGENTS.md           # Symlink to _conduct/AGENTS.md
CLAUDE.md           # Claude instructions (managed block)
WARP.md             # Warp instructions
GEMINI.md           # Gemini instructions

Issue Tracker Integration

Configure GitHub or Linear:

conduct config tracker add
# ? Select tracker: GitHub
# ? Enable immediately? Yes
# ? Default repository: owner/repo

conduct sync --github-token $GITHUB_TOKEN
# 🔄 Syncing with GitHub...
# ✅ Fetched 15 issues
# ✅ Updated 8 connections

JSON Schema Support

Add to conduct.json:

{
  "$schema": "https://git.conduct.run/schema/conduct.json",
  "version": "0.1.0",
  "memory": {
    "profile": "default"
  },
  "trackers": {
    "github": {
      "enabled": true,
      "defaultRepo": "owner/repo"
    }
  }
}

Benefits:

• IDE autocomplete
• Real-time validation
• Property documentation on hover

AI Agent Support

Conduct works with 7+ AI coding assistants:

• Cursor - .cursor/commands/ (9 templates)
• Claude - .claude/commands/ + CLAUDE.md (auto-loaded)
• Warp - .warp/commands/ + WARP.md (Agent Mode)
• Gemini - GEMINI.md instructions
• Factory - .factory/commands/
• OpenCode - .opencode/commands/
• GitHub Copilot - Via AGENTS.md

All agents:

• Share same templates
• Use same memory database
• Generate compatible specs/runs

Example Workflow

# 1. Initialize
conduct init

# 2. Discover existing features
conduct discover -y

# 3. AI agent creates spec (using conduct-spec template)
# → Creates: _conduct/specs/1.v0.spec.md
# → Generates SQL

# 4. Execute SQL
conduct save _conduct/operations/create-spec-1.sql

# 5. AI agent implements (using conduct-run template)
# → Creates: _conduct/runs/1.v0.run.md
# → Links to features

# 6. Execute SQL
conduct save _conduct/operations/update-run-1.sql

# 7. AI agent verifies (using conduct-check template)
# → Creates: _conduct/checks/1.v0.check.md

# 8. Query memory
conduct list

Configuration

Database Profiles

# List profiles
conduct config profile list

# Add new profile
conduct config profile add

# Switch profile
conduct config profile switch production

Tracker Configuration

# Add tracker
conduct config tracker add

# Enable/disable
conduct config tracker enable
conduct config tracker disable

# View configuration
conduct config tracker list

Database Testing

# Test connection
conduct config db test

# Show info
conduct config db info

Requirements

• Node.js: >= 20.0.0
• Operating System: macOS, Linux, Windows

Documentation

• Full Docs: https://conduct.run/docs
• CLI Reference: https://conduct.run/docs/cli-reference
• Getting Started: https://conduct.run/docs/getting-started
• Agent Templates: https://conduct.run/docs/agents
• Architecture: https://conduct.run/docs/architecture

GitHub Repository

https://github.com/21nOrg/conduct

License

MIT - See LICENSE file

Support

• Issues: https://github.com/21nOrg/conduct/issues
• Documentation: https://conduct.run/docs
• Blog: https://conduct.run/blog

Made with ❤️ by 21nOrg