DEV Community

Paarthurnax
Paarthurnax

Posted on

How to Set Up a Crypto AI Agent on Windows in 2026 (Step-by-Step)

#crypto #openclaw #ollama #beginners

How to Set Up a Crypto AI Agent on Windows in 2026 (Step-by-Step)

Running a crypto AI agent on your own Windows PC used to be something only developers with time to burn could pull off. In 2026, that's no longer the case. Tools like OpenClaw and Ollama have made it genuinely beginner-friendly to run a local AI agent that monitors markets, pulls live data from CoinGecko, and paper trades — all without sending your data to a third-party cloud.

This guide walks you through the entire setup, step by step, on Windows.

⚠️ Disclaimer: Nothing in this article is financial advice. Crypto markets are volatile. Paper trading is for learning purposes only. Always do your own research before risking real money.

What You'll Need

Before diving in, make sure you have:

  • Windows 10 or 11 (64-bit)
  • At least 8 GB RAM (16 GB recommended for larger models)
  • ~10 GB free disk space
  • A stable internet connection
  • Basic comfort with PowerShell (no coding required)

Step 1 — Install Node.js

OpenClaw runs on Node.js, so that's your first stop.

  1. Go to https://nodejs.org and download the LTS version for Windows.
  2. Run the installer and accept all defaults.
  3. Open PowerShell (search "PowerShell" in the Start menu) and verify the install: 
node --version
npm --version
Enter fullscreen mode Exit fullscreen mode

You should see version numbers for both. If you do, you're good to move on.

Step 2 — Install OpenClaw

OpenClaw is the AI agent runtime that ties everything together. It handles skills, memory, scheduling, and external tool connections — including crypto data feeds.

Install it globally via npm:

npm install -g openclaw
Enter fullscreen mode Exit fullscreen mode

Once installed, confirm it's working:

openclaw --version
Enter fullscreen mode Exit fullscreen mode

Then initialise your workspace. By default, OpenClaw stores everything in:

C:\Users\<YourUsername>\.openclaw\workspace
Enter fullscreen mode Exit fullscreen mode

Run the gateway to start the local server:

openclaw gateway start
Enter fullscreen mode Exit fullscreen mode

You'll see output confirming the gateway is live. Leave this PowerShell window open, or run it in the background.

Step 3 — Install Ollama for Local AI

Rather than relying on an API key and paying per token, you can run a language model locally on your machine using Ollama. This keeps your trading thoughts private and removes ongoing API costs.

  1. Go to https://ollama.com/download and download the Windows installer.
  2. Run the .exe — it installs to C:\Users\<YourUsername>\AppData\Local\Programs\Ollama by default.
  3. Open a new PowerShell window and pull a model. For a capable but lightweight option: 
ollama pull mistral
Enter fullscreen mode Exit fullscreen mode

This downloads the Mistral 7B model (~4 GB). If you have 16 GB+ RAM, try llama3 for better reasoning:

ollama pull llama3
Enter fullscreen mode Exit fullscreen mode
  1. Test it's running: 
ollama run mistral "What is Bitcoin's halving cycle?"
Enter fullscreen mode Exit fullscreen mode

You should get a coherent response right in your terminal. That's your local AI brain, running entirely on your hardware.

Step 4 — Connect OpenClaw to Ollama

Now we wire OpenClaw to use your local Ollama model instead of a cloud provider.

Open the OpenClaw web UI — by default it runs at:

http://localhost:3000
Enter fullscreen mode Exit fullscreen mode

In the settings panel, find AI Provider and set it to Ollama. Enter the local endpoint:

http://localhost:11434
Enter fullscreen mode Exit fullscreen mode

Select your model (e.g., mistral or llama3) from the dropdown, then save.

From this point on, every agent response runs through your local model — no API keys, no subscriptions, no data leaving your machine.

Step 5 — Set Up CoinGecko Data

CoinGecko offers a free-tier API that gives you real-time and historical crypto prices without requiring a credit card. OpenClaw can query it directly through skills.

Option A — Use the built-in web fetch skill:

OpenClaw's agent can call CoinGecko's public API using the web_fetch tool. No installation needed. Try asking your agent:

What is the current price of Ethereum in USD?
Enter fullscreen mode Exit fullscreen mode

It will fetch https://api.coingecko.com/api/v3/simple/price?ids=ethereum&vs_currencies=usd and return the result.

Option B — CoinGecko API key (optional, higher rate limits):

  1. Sign up free at https://www.coingecko.com/en/api
  2. Copy your API key from the dashboard
  3. In OpenClaw's settings, add an environment variable: 
COINGECKO_API_KEY=your_key_here
Enter fullscreen mode Exit fullscreen mode

With a key, you get significantly higher rate limits — useful if you plan to poll prices frequently.

Step 6 — Configure Your Crypto Agent Persona

OpenClaw uses a SOUL.md file to give your agent personality and focus. Navigate to your workspace:

cd C:\Users\<YourUsername>\.openclaw\workspace
notepad SOUL.md
Enter fullscreen mode Exit fullscreen mode

Add a section tailoring the agent to crypto work:

## Crypto Focus

- Monitor BTC, ETH, and SOL prices daily
- Alert me to >5% price movements
- Summarise market sentiment from on-chain signals
- Never give financial advice — always frame as observation
Enter fullscreen mode Exit fullscreen mode

Save the file. Your agent will load this context at the start of each session.

Step 7 — Run Your First Paper Trade

Paper trading means simulating trades with fake money — it's how you test a strategy without risk.

In the OpenClaw chat interface (at http://localhost:3000), start a conversation with your agent:

Let's paper trade. Assume I have $10,000 USD. 
Check the current prices of BTC and ETH on CoinGecko, 
then suggest an allocation and explain your reasoning.
Enter fullscreen mode Exit fullscreen mode

Your agent will:

  1. Fetch live prices from CoinGecko
  2. Apply basic reasoning from the Ollama model
  3. Suggest a split (e.g., 60% BTC / 40% ETH) with justification

You can then ask follow-up questions:

If BTC drops 10% from here, what's my portfolio value?
Enter fullscreen mode Exit fullscreen mode 
Set a mental stop-loss at $8,500 total. Alert me if we hit it.
Enter fullscreen mode Exit fullscreen mode

OpenClaw can log these paper positions to a file in your workspace under memory/ — giving you a running ledger of your simulated trades across sessions.

Step 8 — Schedule Daily Market Briefings

One of the most useful things you can do is have your agent brief you every morning automatically.

In your workspace, open or create HEARTBEAT.md:

notepad C:\Users\<YourUsername>\.openclaw\workspace\HEARTBEAT.md
Enter fullscreen mode Exit fullscreen mode

Add:

## Daily Crypto Brief

Every morning, fetch BTC, ETH, and SOL prices from CoinGecko.
Compare to yesterday's close. Note any >3% moves.
Summarise in 3 bullet points. Log to memory/crypto-brief-YYYY-MM-DD.md.
Enter fullscreen mode Exit fullscreen mode

When OpenClaw's heartbeat triggers (configurable in settings, e.g., every 30 minutes or at a set cron time), your agent will run this automatically and save the output.

Troubleshooting Common Issues

Ollama not responding:

ollama serve
Enter fullscreen mode Exit fullscreen mode

Run this if the Ollama daemon stopped. It restarts the local server on port 11434.

OpenClaw gateway not starting:

openclaw gateway status
openclaw gateway restart
Enter fullscreen mode Exit fullscreen mode

CoinGecko returning 429 (rate limited):
Wait 60 seconds and retry, or add a free API key (Step 5, Option B). The free tier allows ~30 calls/minute without a key.

Node version too old:
OpenClaw requires Node 18+. Update via https://nodejs.org or use nvm-windows to manage versions.

What's Next

Once your agent is running and paper trading comfortably, you can extend it further:

  • Add more data sources — DeFiLlama for TVL data, Fear & Greed Index, on-chain metrics
  • Build custom skills — OpenClaw's skill system lets you write small tools in plain JavaScript
  • Upgrade your model — swap Mistral for a larger model when your hardware supports it
  • Automate alerts — connect to Discord or email to get notified when thresholds are hit

The foundation you've built here — local AI, live data, structured memory — is exactly what professional quant traders pay thousands for in cloud subscriptions. You're running it on your own hardware, on Windows, for free.

Want a Ready-Made Setup?

If you'd rather skip the manual configuration and get a pre-built crypto agent kit with skills, prompts, and templates already wired together, check out the Home AI Agent bundle:

👉 https://dragonwhisper36.gumroad.com/l/homeaiagent

It includes everything covered in this guide — plus extras like pre-configured market briefing templates, paper trade logging sheets, and a step-by-step setup video.

⚠️ Reminder: This article is for educational purposes only. Nothing here constitutes financial advice. Crypto assets are highly volatile and you can lose money. Always consult a qualified financial adviser before making investment decisions.

Top comments (0)