← All posts

Getting Started: Setting Up Your AI Agent

Two options for powering your OpenClaw agent's AI brain — a budget-friendly flexible option and a premium one. Pick what fits.

Client setup guide for OpenClaw AI agents

Option A: OpenRouter

OpenRouter is a single gateway that gives your agent access to 300+ AI models from every major provider (Google, Anthropic, OpenAI, and more) through one account and one API key. You can start cheap and upgrade models anytime without changing anything on the server.

Why OpenRouter?

  • One account, all models — switch between cheap and premium models anytime
  • Pay-as-you-go — prepaid credits, no surprise bills
  • Budget controls — set spending limits per API key
  • Best value — start with Gemini Flash at ~$0.50/$3 per million tokens, upgrade to Claude Sonnet only if needed

Step 1: Create an OpenRouter Account

  1. Go to openrouter.ai
  2. Click Sign Up (top right) — you can use email or Google
  3. Verify your email if prompted

Step 2: Add Credits

  1. Once logged in, go to openrouter.ai/credits
  2. Add credits via credit card (minimum $5)
  3. I recommend starting with $20 — this will last a while on budget models

Credits don't expire and are deducted automatically as the agent uses them. You can top up anytime.

Step 3: Create an API Key

  1. Go to openrouter.ai/settings/keys
  2. Click Create Key
  3. Name it something like "OpenClaw Agent"
  4. Set a credit limit — I recommend $20/month to start as a safety net
  5. Copy the key immediately — it starts with sk-or- and won't be shown again

Step 4: Send Me the Key

I'll configure your agent with the key securely on your server. After setup, you can monitor your usage at openrouter.ai/activity.


What Model Will My Agent Use?

Your agent will start on Google Gemini Flash, which handles most everyday agent tasks well and is very affordable. Here's how the models compare:

Model Cost (per 1M tokens in/out) Monthly Estimate* Best For
Gemini 3 Flash $0.50 / $3.00 ~$5–15/mo Daily tasks, conversation, research
Claude Haiku 4.5 $1.00 / $5.00 ~$10–25/mo Better reasoning, still affordable
Claude Sonnet 4.6 $3.00 / $15.00 ~$30–90/mo Complex tasks, best quality
DeepSeek V3.2 $0.14 / $0.28 ~$2–5/mo Cheapest capable option

*Estimates based on moderate agent usage (research, messaging, scheduling tasks). Actual costs depend on how actively the agent is used.

The bottom line: Gemini Flash gives you 80–90% of the quality at roughly 10% of the cost of premium models. If we find it's not handling certain tasks well enough, we can upgrade to Haiku or Sonnet anytime — it's just a config change on the server, no action needed from you.


Option B: Direct Anthropic API

If you want the best possible AI quality and don't mind higher costs, you can go directly with Anthropic's Claude. This is the model OpenClaw was originally built for, so tool use and instruction following are top-notch.

Step 1: Create an Anthropic Account

  1. Go to console.anthropic.com
  2. Click Sign Up
  3. Create an account with your email (or sign in with Google)

Step 2: Add a Payment Method

  1. Once logged in, click Settings in the left sidebar
  2. Go to Billing
  3. Click Add Payment Method and enter a credit card
  4. Set a monthly spending limit — I recommend starting with $50/month as a safety net

The spending limit prevents unexpected charges. If the agent hits the limit, it pauses until the next billing cycle or until you raise the limit.

Step 3: Create an API Key

  1. In the left sidebar, click API Keys
  2. Click Create Key
  3. Give it a name like "OpenClaw Agent"
  4. Copy the key immediately — it starts with sk-ant- and won't be shown again

Step 4: Send Me the Key

Same as above — I'll configure it securely on the server.

Anthropic Pricing

Model Cost (per 1M tokens in/out) Monthly Estimate*
Claude Sonnet 4.6 $3.00 / $15.00 ~$30–90/mo
Claude Haiku 4.5 $1.00 / $5.00 ~$10–25/mo
Claude Opus 4.6 $5.00 / $25.00 ~$50–150/mo

My Recommendation

Start with Option A (OpenRouter) using Gemini Flash. It's the most cost-effective way to get up and running. Your agent will handle daily tasks well, and we can always upgrade the model later if needed. The switch takes minutes and requires no effort on your end.

If you're seeing costs you're uncomfortable with or want better quality for specific tasks, we can discuss upgrading — it's always just a model switch away.


Sending the Key Securely

Please don't send API keys over regular email or text. Use one of these:

  • One-time secret link: Go to onetimesecret.com, paste the key, and send me the generated link
  • Signal or WhatsApp disappearing messages
  • We can do it together on a screen share — I'll walk you through pasting it directly into the config

Questions?

Reach out anytime. Once I have your API key, I'll have your agent configured and running within the hour.

Email us Join our Discord