Skip to main content

​
What You’ll Build

In this guide, you’ll create a hub-mode workspace from scratch — complete with memory, skills, and multiple agents configured. By the end, you’ll have a self-evolving workspace that works with any AI coding agent.

​
Prerequisites

  • Node.js ≥ 18
  • At least one AI coding agent installed (e.g., GitHub Copilot, Claude Code)

​
Step 1: Create the Workspace

mkdir my-ai-workspace && cd my-ai-workspace
npx drevon init
The wizard starts. Here’s what to choose:
1

Mode → Hub

Since this is a new empty directory, Drevon suggests Hub mode. Accept it.
2

Name → my-ai-workspace

Uses the directory name. Press Enter to accept.
3

Identity → Founder

Choose “Founder” for maximum autonomy. The agent will take initiative and move fast.
4

Agents → Copilot + Claude

Select at least two agents to see Drevon’s multi-agent power.
5

Memory → Yes

Enable memory for persistent cross-session context.
6

Starter Prompts → Yes

Install the 5 starter workflow templates.

​
Step 2: Explore What Was Created

drevon status
You should see:
  • Mode: hub
  • Identity: founder-agent
  • 2 agents enabled with health checks
  • 4 memory files
  • 1 skill (find-skills)
  • 5 prompts
Let’s look at the structure: 
ls -la

drevon.config.json
.drevon/
  memory/
    user.md
    projects.md
    systems.md
    log.md
  skills/
    find-skills/
  prompts/
    _index.md
    code-review.md
    debug.md
    new-feature.md
    refactor.md
    write-tests.md
workspace/
.github/copilot-instructions.md
CLAUDE.md

​
Step 3: Install Your First Skill

Let’s add browser automation capability: 
drevon skill add browser-use/browser-use
Verify it’s installed: 
drevon skill list

​
Step 4: Add a Custom Instruction

Edit drevon.config.json to add a project-specific rule: 
{
  "instructions": [
    {
      "id": "memory-protocol",
      "title": "Memory Protocol",
      "content": "Read all files in .drevon/memory/ at the start of every session. Write to them after significant actions.",
      "alwaysApply": true
    },
    {
      "id": "tech-stack",
      "title": "Tech Stack",
      "content": "This workspace uses TypeScript, Next.js 15, Tailwind CSS, and Vercel for deployment. Always use these technologies unless explicitly told otherwise.",
      "alwaysApply": true
    }
  ]
}
Then sync: 
drevon sync

​
Step 5: Start Your First Project

Create a project folder: 
mkdir workspace/my-first-project
Now open the workspace in your editor and start a conversation with your AI agent. The agent will:
  1. Read memory files to understand the workspace context
  2. Follow your identity rules (founder-agent posture)
  3. Know about installed skills (browser-use, find-skills)
  4. Use your custom instructions (tech stack preferences)

​
Step 6: Verify the Self-Evolution Loop

After your first session, check if the agent updated memory: 
cat .drevon/memory/log.md
You should see new entries documenting what the agent did. Over time, these memory files become an invaluable knowledge base.

​
What’s Next

Add More Skills

Browse skills.sh for more capabilities.

Custom Instructions

Write powerful project-specific rules.

Team Setup

Share this workspace with your team.

Best Practices

Tips for getting the most out of Drevon.