Documentation
Getting Started

Getting Started

Conduct is an AI agent orchestration platform with persistent memory that enables AI agents to work with structured workflows, automated codebase understanding, and intelligent memory management. This guide will help you get started in minutes.

Installation

Choose your preferred installation method:

via install script (Recommended) 

curl -fsSL https://git.conduct.run/install.sh | bash

via npm 

npm install -g conduct-cli

via npx (No Installation) 

npx conduct-cli init

From Source 

git clone https://github.com/21nOrg/conduct
cd conduct
pnpm install
cd cli
npm run build
npm link

Core Concepts

Spec → Run → Check Workflow

Conduct implements a three-phase workflow with memory persistence:

User Intent → Specification → Implementation → Verification
     ↓               ↓              ↓              ↓
Memory Database  Memory Database  Memory Database  Memory Database

Separation of Concerns:

  • Spec: What to build (intent, requirements, approach)
  • Run: How it was built (execution log, decisions, changes)
  • Check: Verification results (pass/fail, gaps, issues)

Persistent Memory

All work is stored in a libSQL memory database that remembers:

  • ✅ Features in your codebase (discovered automatically)
  • ✅ Specifications and execution runs
  • ✅ Verification checks and results
  • ✅ Relationships between features and runs
  • ✅ Remote issue connections (GitHub, Linear)

Initialize Your Project

Run the initialization command in your project directory:

conduct init

This interactive command will:

  1. Choose Database Mode: Select local (embedded) or remote (Turso) database
  2. Configure Profile: Set up database credentials
  3. Create Directory Structure: Initialize _conduct/ workspace
  4. Create Agent Instructions: Generate AGENTS.md, CLAUDE.md, WARP.md, GEMINI.md
  5. Install Agent Templates: Deploy 9 agent command templates to 6+ AI assistants

What Gets Created

After initialization, Conduct creates:

your-project/
├── _conduct/                   # Conduct working directory
│   ├── specs/                  # Specifications
│   ├── runs/                   # Execution logs
│   ├── checks/                 # Verification reports
│   ├── designs/                # UI designs
│   ├── operations/             # SQL files for memory ops
│   ├── templates/              # Agent command templates
│   ├── logs/                   # System logs
│   ├── memory.db              # Local libSQL database
│   └── .meta/                 # AI-generated docs
├── conduct.json               # Configuration
├── conduct.track.json         # Fast local tracker
├── AGENTS.md                  # Symlink to _conduct/AGENTS.md
├── CLAUDE.md                  # Claude-specific instructions
├── WARP.md                    # Warp-specific instructions
├── GEMINI.md                  # Gemini-specific instructions
├── .cursor/commands/          # Cursor agent templates
├── .claude/commands/          # Claude agent templates
├── .warp/commands/            # Warp agent templates
├── .factory/commands/         # Factory agent templates
├── .opencode/commands/        # OpenCode agent templates
└── ~/.conduct/credentials     # Global credentials file

First Workflow

Here’s a complete example workflow:

1. Discover Features

Automatically discover features in your codebase:

conduct discover -y

This scans your codebase and identifies features using:

  • Directory structure analysis
  • Export analysis (TypeScript/JavaScript)
  • package.json hints

2. Create a Specification

Use the AI agent with the conduct-spec template to create a specification:

User: "Build user authentication with OAuth2 support"
Agent: (uses conduct-spec template)
  - Fetches remote context (GitHub issues, Linear tickets)
  - Consults memory (past specs, existing features)
  - Creates _conduct/specs/1.v0.spec.md
  - Generates SQL to register in memory

3. Execute the Specification

Use the AI agent with the conduct-run template:

Agent: (uses conduct-run template)
  - Checks for UI designs first
  - Links to discovered features
  - Implements specification end-to-end
  - Creates _conduct/runs/1.v0.run.md
  - Generates SQL to log execution

4. Verify Implementation

Use the AI agent with the conduct-check template:

Agent: (uses conduct-check template)
  - Compares spec vs. run
  - Verifies code vs. spec
  - Detects gaps
  - Creates _conduct/checks/1.v0.check.md
  - Generates SQL to log results

5. Update Memory

Review and execute the generated SQL:

# Review the SQL
cat _conduct/operations/update-run-1.sql

# Validate and execute
conduct save _conduct/operations/update-run-1.sql

Query Your Memory

List everything in memory:

conduct list                # Show everything
conduct list --specs        # Show only specs
conduct list --runs         # Show only runs
conduct list --features     # Show only features
conduct list --checks       # Show only checks
conduct list --json         # Output as JSON

Example output:

📋 Specs (3):
  spec-1 [completed] Authentication System (simple)
  spec-2 [in-progress] Dashboard UI (medium)
  spec-3 [pending] API Integration (complex)

🏃 Runs (2):
  1 [completed] spec-1.v0 - Implemented OAuth2 auth
  2 [in-progress] spec-2.v0 - Building dashboard

✨ Features (15):
  #1 authentication - Authentication System [active]
  #2 user-profile - User Profile Management [active]
  #3 dashboard - Analytics Dashboard [active]

Maintenance

Weekly Tasks 

conduct relevancy --recalculate-all  # Update time decay
conduct archive -y                   # Remove old runs
conduct reconcile                    # Detect drift
conduct health                       # Check system

Monthly Tasks 

conduct reconcile --since-days 30 -y # Full reconciliation
conduct discover -y                  # Discover new features
conduct archive --threshold 0.15 -y  # Aggressive archiving

Upgrade Instructions

Keep your agent templates up to date:

conduct upgrade

This updates all agent command templates to the latest version.

Next Steps

CLI Reference