Install Eva-01

Step-by-Step

The Installation Process

7 steps. Copy the commands, follow the prompts, meet your new Chief of Staff.

Install OpenClaw

OpenClaw is the engine that powers Eva-01. One command installs it.

npm install -g openclaw

Don't have Node.js? Download it from nodejs.org first (LTS version). Then come back and run this command.

Official OpenClaw install guide

Unzip Your Eva-01 Blueprint

Unzip the file you downloaded. Move it somewhere permanent.

mv ~/Downloads/*eva-01*blueprint* ~/eva-workspace
cd ~/eva-workspace
ls

You should see files like AGENTS.md, SOUL.md, USER.md, MEMORY.md, and others. If you see them, you're in the right place.

Point OpenClaw at Your Blueprint

Tell OpenClaw where your Eva-01 files live.

openclaw onboard --install-daemon

The wizard will ask you to choose a model provider and set an API key. Follow the prompts. When it asks for a workspace, point it to the folder you just unzipped.

Add Your API Key

Eva-01 needs an AI model to think. Anthropic's Claude is recommended.

1 Go to console.anthropic.com
2 Sign up and add a payment method
3 Create an API key (Settings → API Keys → Create Key)
4 The onboarding wizard will ask for it. Paste it in.

Already use Claude Desktop, Codex, or Cursor? OpenClaw imports your settings automatically. You might not even need this step.

Connect a Messaging Channel

This is how you'll talk to Eva-01 from your phone or desktop.

openclaw channel add telegram

Follow the prompts. You'll need to create a Telegram bot via @BotFather. It takes about 2 minutes. You can also use Discord, WhatsApp, Signal, or just the built-in chat.

Send Your First Prompt (This Is the Important One)

Before you start chatting normally, send this first message. It tells your new agent to adopt the Eva-01 framework as its operating system.

I just set up your workspace with the Eva-01 framework from evaonline.xyz. Before we start working together, I need you to do a complete onboarding:

1. Read every .md file in your workspace root, in this order:
   - IDENTITY.md (who you are)
   - SOUL.md (your personality, voice, and how to handle my emotional states)
   - USER.md (who I am, my full profile)
   - AGENTS.md (your operating rules, security directives, delegation rules, learning system)
   - HEARTBEAT.md (what to check periodically)
   - CURRENT-PRIORITIES.md (what I'm working on this week)
   - MEMORY.md (your long-term memory, empty for now)
   - RESEARCH.md (your research log, empty for now)
   - TOOLS.md (environment notes)

2. After reading all files, tell me:
   - My name and what you know about me
   - Your name and how you'd describe your personality
   - My top 3 priorities
   - The security rules you'll follow (Constitutional Directives)
   - How you'll learn about me over time

3. Create today's daily log file at memory/YYYY-MM-DD.md with this template:
   # Daily Note — [today's date] ([day of week])
   
   ## What Happened
   - First boot. Eva-01 framework adopted.
   
   ## Action Items
   - [ ] Owner to fill in any missing USER.md sections
   
   ## What I Learned
   - [Write your first observations about me based on my profile]
   
   ## Carrying Forward
   - [ ] Get to know the owner through conversation

4. If any file is missing or empty, tell me what's missing so I can fix it.

This is our foundation. Everything builds from here. Go ahead.

That's it. Your agent will read all the files, adopt the Eva-01 personality, and greet you as your new Chief of Staff. From here on, just talk to her like a person.

Verify Your Setup

After your agent finishes the onboarding, test a few things.

Quick check:
1. What are my top 3 priorities right now?
2. What's my communication preference?
3. If I seem overwhelmed, what would you do?
4. What data classification tier is my health information?

If your agent answers all four correctly, you're fully set up. If anything seems off, re-check that USER.md and SOUL.md are in your workspace.

Troubleshooting

Common Issues

Most problems have a one-line fix. Check here first.

"command not found" ▼

Close and reopen your terminal. This refreshes your PATH so it can find the newly installed openclaw command. If it still doesn't work, make sure Node.js is installed: run node --version to check.

"Invalid API key" ▼

Double-check your API key at console.anthropic.com. Make sure you copied the full key (it starts with sk-ant-), and that you have billing set up on your Anthropic account.

"Eva seems generic" ▼

Make sure USER.md and SOUL.md are in your workspace root (the same folder you pointed OpenClaw at). If they're inside a subfolder, your agent can't find them. Move them to the top level and re-send the first prompt from Step 6.

"Agent didn't read my files" ▼

Re-send the first prompt from Step 6. Sometimes the agent needs an explicit instruction to read all workspace files. Copy it exactly from above, including all 4 numbered sections.

"I want to start over" ▼

Delete the contents of the memory/ folder (but keep the folder itself). Then re-send the first prompt from Step 6. Your agent will re-read all files and start fresh, as if it's day one again.

What You'll Need