OpenClaw + ClaudeAPI.com: Complete Setup Guide (About 5 Minutes)

OpenClaw is a popular open-source personal AI assistant framework (GitHub shows 60,000+ stars at the time of writing). It connects Claude to everyday chat surfaces such as Telegram, WhatsApp, Discord, and Slack, so you can talk to your model from the apps you already use.

This walkthrough shows how to configure OpenClaw end-to-end using an API key from ClaudeAPI.com.

1. What is OpenClaw?

OpenClaw runs on your own machine or server and acts as a local AI gateway:

• Receives messages from Telegram, WhatsApp, Discord, and similar channels
• Forwards them to Claude using your API key
• Sends the model’s replies back to you

Typical benefits: you keep control of your data plane, you can extend behavior with skills, and you get a single gateway across channels. You bring your own key and pay per usage according to your ClaudeAPI account.

2. Prerequisites

Check Node.js

node --version
# Expect v22.x.x or v24.x.x

node --version
# Expect v22.x.x or v24.x.x

If the version is below 22, upgrade—for example with nvm:

nvm install --lts && nvm use --lts

nvm install --lts && nvm use --lts

Get a ClaudeAPI.com API key

1. Open claudeapi.com and sign up or sign in.
2. Go to the dashboard → API Keys → create a new key.
3. Copy and store it securely (keys are often shown with an sk-ant- style prefix—use whatever format your dashboard displays).

3. Install OpenClaw

macOS / Linux (recommended one-liner)

curl -fsSL https://openclaw.ai/install.sh | bash

curl -fsSL https://openclaw.ai/install.sh | bash

Windows (PowerShell as Administrator)

iwr -useb https://openclaw.ai/install.ps1 | iex

iwr -useb https://openclaw.ai/install.ps1 | iex

Alternative: global npm install

npm install -g openclaw@latest

npm install -g openclaw@latest

Verify

openclaw --version

openclaw --version

If you see a version string, the CLI is installed.

4. Run the onboarding wizard

openclaw onboard --install-daemon

openclaw onboard --install-daemon

--install-daemon registers the OpenClaw gateway as a background service so it can start automatically after reboots—recommended.

Wizard highlights

Step 1: AI provider

You may see a prompt similar to:

Which AI provider do you want to use?
> Anthropic (Claude) — recommended
  OpenAI (ChatGPT)
  Custom provider

Which AI provider do you want to use?
> Anthropic (Claude) — recommended
  OpenAI (ChatGPT)
  Custom provider

Choose Custom.

Step 2: API key

Paste your ClaudeAPI.com API key when prompted:

Enter your API key:
> (paste your ClaudeAPI.com key, then press Enter)

Enter your API key:
> (paste your ClaudeAPI.com key, then press Enter)

Step 3: Default model

Example list:

Select your default model:
> claude-sonnet-4-6    ← balanced cost/quality
  claude-opus-4-7      ← strongest, heavier tasks
  claude-haiku-4-5     ← fastest, simple Q&A

Select your default model:
> claude-sonnet-4-6    ← balanced cost/quality
  claude-opus-4-7      ← strongest, heavier tasks
  claude-haiku-4-5     ← fastest, simple Q&A

New users often start with claude-sonnet-4-6.

Step 4: Set ClaudeAPI.com baseUrl (critical)

The wizard defaults to sending requests to Anthropic’s official servers. To use ClaudeAPI.com as the forwarder, you must set baseUrl accordingly (in the wizard when offered, or in config afterward).

Important: baseUrl must be https://gw.claudeapi.com—the correct ClaudeAPI.com gateway address (not api.claudeapi.com).

Per the summary table, you can also confirm this value in ~/.openclaw/openclaw.json after onboarding.

Step 5: Install the daemon

Confirm when the wizard offers to install the background service.

Restart the gateway so config applies

# macOS (launchd)
launchctl stop ai.openclaw.gateway
launchctl start ai.openclaw.gateway

# Linux (systemd user unit)
systemctl --user restart openclaw-gateway

# Or use the CLI
openclaw gateway restart

# macOS (launchd)
launchctl stop ai.openclaw.gateway
launchctl start ai.openclaw.gateway

# Linux (systemd user unit)
systemctl --user restart openclaw-gateway

# Or use the CLI
openclaw gateway restart

5. Verify everything works

Option A: CLI smoke test

openclaw chat "Hello — please introduce yourself briefly."

openclaw chat "Hello — please introduce yourself briefly."

If you get a normal Claude-style reply, routing is working.

Option B: Dashboard

openclaw dashboard

openclaw dashboard

A browser should open the local control panel (commonly http://127.0.0.1:18789). Send a message in the UI; a reply confirms the stack is up.

Option C: Model status

openclaw models status

openclaw models status

You should see the Anthropic-related provider in an active state (exact labels depend on OpenClaw version).

6. Models available via ClaudeAPI.com

In config, agents.defaults.model.primary can reference models with an anthropic/ prefix, for example:

Pricing shown here matches the source article; always confirm current rates in your ClaudeAPI dashboard before budgeting.

7. Troubleshooting

ANTHROPIC_API_KEY not set

The key was not picked up. Check:

# Option A: environment variable
export ANTHROPIC_API_KEY="sk-ant-..."

# Option A: environment variable
export ANTHROPIC_API_KEY="sk-ant-..."

Option B: edit ~/.openclaw/agents/main/agent/auth-profiles.json (path may vary):

{
  "version": 1,
  "profiles": {
    "anthropic:default": {
      "type": "api_key",
      "provider": "anthropic",
      "key": "sk-ant-YOUR_CLAUDEAPI_KEY"
    }
  }
}

{
  "version": 1,
  "profiles": {
    "anthropic:default": {
      "type": "api_key",
      "provider": "anthropic",
      "key": "sk-ant-YOUR_CLAUDEAPI_KEY"
    }
  }
}

HTTP 401 / “Invalid API key”

Double-check baseUrl is exactly https://gw.claudeapi.com and that the key is copied from the ClaudeAPI dashboard without extra spaces.

Windows install issues

Prefer WSL2. In PowerShell: wsl --install, then follow the Linux steps inside WSL.

Gateway exits immediately

Run openclaw doctor --fix and inspect logs. Confirm port 18789 is not already in use by another process.

8. Next step: skills

OpenClaw supports skills (extensions) such as web search, email, or calendar helpers:

openclaw skills list
openclaw skills install web-search

openclaw skills list
openclaw skills install web-search

Summary

After that, you have a locally hosted assistant that reaches Claude through ClaudeAPI.com, ready to wire to Telegram, Discord, and other channels per OpenClaw’s channel docs.

For more guides or account help, see claudeapi.com.

Suggested internal links (for editors)

• ClaudeAPI getting-started / API key documentation (English)
• Model list and pricing page aligned with the IDs in this article
• OpenAI-compatible vs Anthropic-native endpoints (if you maintain a comparison page)

CTA

Create a key in the ClaudeAPI dashboard, set baseUrl to https://gw.claudeapi.com, finish openclaw onboard, then connect your first chat channel using OpenClaw’s channel guides.