OpenClaw Manual Installation - Smart Model Routing

Overview

OpenClaw is an open-source AI agent Gateway that acts as a bridge between chat applications and AI agents. Through a centralized Gateway process, it can connect chat platforms like Telegram, WhatsApp, Discord, and Feishu to AI programming agents. This document describes how to manually install OpenClaw and configure EvoLink API as a model provider. After completing this document, you can continue to configure specific chat channels (such as Telegram or Feishu). This guide covers:

Installing and configuring OpenClaw Gateway
Configuring EvoLink API as a custom model provider
Verifying the installation

System Environment Check (Optional)

Before starting the installation, it’s recommended to run the environment check tool to ensure your system meets OpenClaw’s requirements.

Download the Check Tool

Download the check tool for your platform from GitHub Releases:

Platform	Filename
Windows	`openclaw-checker-win-x64.exe`
macOS (Intel)	`openclaw-checker-macos-x64`
macOS (Apple Silicon)	`openclaw-checker-macos-arm64`
Linux	`openclaw-checker-linux-x64`

Check Items

The tool will automatically check the following:

✅ Node.js version (requires >= 22.12.0)
✅ npm available
✅ Git available
✅ Network connectivity (github.com, npmjs.org, evolink.ai)

If the check fails, the tool will provide specific fix suggestions.

Prerequisites

Before starting configuration, ensure you have completed the following:

1. Install Node.js

OpenClaw is installed via npm and requires Node.js 22 or higher.

Windows
macOS

Visit Node.js official website, download the Windows installer (.msi file), and run the installation program.After installation, open PowerShell to verify:

node --version
npm --version

It’s recommended to run PowerShell as administrator to avoid permission issues during installation.

Method 1: Using InstallerVisit Node.js official website, download the macOS installer (.pkg file), and run the installation program.Method 2: Using Homebrew

brew install node

After installation, open Terminal to verify:

node --version
npm --version

If you encounter permission issues during installation, you may need to add sudo before the command.

2. Get EvoLink API Key

Log in to EvoLink Console
Find API Keys in the console, click “Create New Key”, then copy the generated Key
API Key usually starts with sk-, please keep it safe

Step 1: Install OpenClaw

Execute in terminal:

npm install -g openclaw@latest

Verify after installation:

openclaw --version

Step 2: Initialize Setup

Run the onboarding command, OpenClaw will guide you through initial configuration and install the daemon service:

openclaw onboard --install-daemon

1. Confirm Installation

The system will prompt installation risk notice, confirm to continue:

2. Select Installation Mode

The system will prompt to select installation mode, choose Quickstart:

3. Select Provider

The system will prompt to select model provider, choose Skip here, we will manually configure EvoLink as a custom provider later:

4. Select Models

The system will prompt to select models to enable, choose All:

5. Select Default Model

The system will prompt to select default model, choose Keep current:

6. Select Channel

The system will prompt to select a chat channel. It’s recommended to choose Skip for now, you can add channels later:

7. Configure Skills

The system will prompt whether to configure Skills. It’s recommended to choose No, you can add them later:

8. Enable Hooks

The system will prompt whether to enable Hooks. It’s recommended to choose session-memory:

9. Restart Gateway Service

The system will prompt that the gateway service is already installed, choose Restart:

10. Launch Bot

The system will prompt how to launch the bot. It’s recommended to choose Do this later:

Step 3: Configure EvoLink API & Model Switching

1. Locate Two Configuration Files (Important)

OpenClaw model configuration typically involves two files:

Windows
macOS

openclaw.json: %USERPROFILE%\.openclaw\openclaw.json
models.json: %USERPROFILE%\.openclaw\agents\main\agent\models.json

openclaw.json: ~/.openclaw/openclaw.json
models.json: ~/.openclaw/agents/main/agent/models.json

Open directly:

open ~/.openclaw/openclaw.json
open ~/.openclaw/agents/main/agent/models.json

If a provider’s apiKey / baseUrl in models.json is non-empty, it will override the corresponding values in openclaw.json. It’s recommended to keep both consistent.

2. Configure Model Providers

It’s recommended to configure the following providers in openclaw.json (and sync to models.json):

"models": {
  "providers": {
    "evolink-anthropic": {
      "api": "anthropic-messages",
      "baseUrl": "https://direct.evolink.ai",
      "apiKey": "Your EvoLink API Key",
      "models": [
        { "id": "evolink/auto", "name": "EvoLink Auto" },
        { "id": "claude-opus-4-6", "name": "Claude Opus 4.6" },
        { "id": "claude-sonnet-4-6", "name": "Claude Sonnet 4.6" },
        { "id": "claude-opus-4-5-20251101", "name": "Claude Opus 4.5" },
        { "id": "claude-opus-4-1-20250805", "name": "Claude Opus 4.1" },
        { "id": "claude-sonnet-4-5-20250929", "name": "Claude Sonnet 4.5" },
        { "id": "claude-sonnet-4-20250514", "name": "Claude Sonnet 4" },
        { "id": "claude-haiku-4-5-20251001", "name": "Claude Haiku 4.5" }
      ]
    },
    "evolink-google": {
      "api": "google-generative-ai",
      "baseUrl": "https://direct.evolink.ai/v1beta",
      "apiKey": "Your EvoLink API Key",
      "models": [
        { "id": "evolink/auto", "name": "EvoLink Auto" },
        { "id": "gemini-3.1-flash-lite-preview", "name": "Gemini 3.1 Flash Lite" },
        { "id": "gemini-3.1-pro-preview", "name": "Gemini 3.1 Pro" },
        { "id": "gemini-2.5-pro", "name": "Gemini 2.5 Pro" },
        { "id": "gemini-2.5-flash", "name": "Gemini 2.5 Flash" },
        { "id": "gemini-3-pro-preview", "name": "Gemini 3.0 Pro" },
        { "id": "gemini-3-flash-preview", "name": "Gemini 3.0 Flash" }
      ]
    },
    "evolink-openai": {
      "api": "openai-completions",
      "baseUrl": "https://direct.evolink.ai/v1",
      "apiKey": "Your EvoLink API Key",
      "models": [
        { "id": "gpt-5.4", "name": "GPT-5.4" },
        { "id": "gpt-5.2", "name": "GPT-5.2" },
        { "id": "gpt-5.1", "name": "GPT-5.1" },
        { "id": "gpt-5.1-chat", "name": "GPT-5.1 Chat" },
        { "id": "gpt-5.1-thinking", "name": "GPT-5.1 Thinking" },
        { "id": "gemini-2.5-pro", "name": "Gemini 2.5 Pro (OpenAI SDK)" },
        { "id": "gemini-2.5-flash", "name": "Gemini 2.5 Flash (OpenAI SDK)" },
        { "id": "gemini-3-pro-preview", "name": "Gemini 3.0 Pro (OpenAI SDK)" },
        { "id": "gemini-3-flash-preview", "name": "Gemini 3.0 Flash (OpenAI SDK)" },
        { "id": "doubao-seed-2.0-pro", "name": "Doubao Seed 2.0 Pro" },
        { "id": "doubao-seed-2.0-lite", "name": "Doubao Seed 2.0 Lite" },
        { "id": "doubao-seed-2.0-mini", "name": "Doubao Seed 2.0 Mini" },
        { "id": "doubao-seed-2.0-code", "name": "Doubao Seed 2.0 Code" },
        { "id": "kimi-k2-thinking", "name": "Kimi K2 Thinking" },
        { "id": "kimi-k2-thinking-turbo", "name": "Kimi K2 Thinking Turbo" }
      ]
    }
  }
}

The model IDs above are examples. Please use the models actually available in your EvoLink account.

For Gemini scenarios, evolink-google.baseUrl must include /v1beta, i.e., https://direct.evolink.ai/v1beta. Without this suffix, you may encounter Forbidden (403) errors.

3. Configure Default Model (Supports Quick Switching)

Set the default model in agents.defaults. We recommend using Smart Model Routing evolink/auto, which automatically selects a suitable model based on your request:

"agents": {
  "defaults": {
    "model": {
      "primary": "evolink-anthropic/evolink/auto"
    }
  }
}

Smart Model Routing (EvoLink Auto): Use evolink/auto as the model ID, and the system will automatically select a suitable model from the model pool based on request complexity, length, and type. No manual switching needed — ideal for most general-purpose scenarios. See EvoLink Auto Documentation for details.

To specify a particular model, you can also switch manually:

Smart Routing: evolink-anthropic/evolink/auto (Recommended)
Claude: evolink-anthropic/claude-opus-4-6
GPT: evolink-openai/gpt-5.2
Gemini: evolink-google/gemini-3.1-pro-preview
Doubao: evolink-openai/doubao-seed-2.0-mini

4. Quick Switch to EvoLink Models (Recommended)

After completing provider configuration, it’s recommended to use CLI commands for model switching instead of manually editing JSON:

# View configured EvoLink OpenAI-compatible models
openclaw models list --provider evolink-openai --plain

# Switch default model (example: gpt-5.4)
openclaw models set evolink-openai/gpt-5.4

# View current active model
openclaw models status --plain

If models list --provider evolink-openai doesn’t show your expected models, check whether both openclaw.json and models.json have the corresponding provider configured.

5. Restart and Verify

Restart the gateway after configuration:

openclaw gateway restart

Check status:

openclaw gateway status

Send a test message to verify the model is working:

openclaw agent --agent main -m "hi" --json

Common Commands

Command	Description
`openclaw gateway status`	Check gateway running status
`openclaw gateway restart`	Restart gateway service
`openclaw gateway stop`	Stop gateway service
`openclaw gateway start`	Start gateway service
`openclaw logs --follow`	View gateway logs in real-time
`openclaw plugins list`	View installed plugins

Troubleshooting

Issue	Solution
npm installation fails	Windows: Run PowerShell as administrator; macOS: Add `sudo` before command
Configuration file not found	Confirm onboard process is complete, check if `~/.openclaw/` directory exists
Gateway fails to start	Check if port is occupied, use `openclaw gateway status` to view detailed errors
Invalid API Key	Confirm API Key is copied correctly, check for extra spaces or quotes
Model configuration not effective	Check both `openclaw.json` and `models.json` for consistency (`models.json` may override)
Gemini returns `Forbidden (403)`	Check if `models.providers.evolink-google.baseUrl` is `https://direct.evolink.ai/v1beta` (must include `/v1beta`)
Old model still used after switching	Run `openclaw models status --plain` to confirm current model, restart with `openclaw gateway restart` if necessary

Next Steps

OpenClaw installation and EvoLink API configuration are complete. Next you can:

Configure Telegram Channel: Refer to OpenClaw + Telegram documentation
Configure Feishu Channel: Refer to OpenClaw + Feishu documentation
Use Auto-Installation Tool: Refer to OpenClaw Auto Installation documentation

AI Coding Tools

​Overview

​System Environment Check (Optional)

​Download the Check Tool

​Check Items

​Prerequisites

​1. Install Node.js

​2. Get EvoLink API Key

​Step 1: Install OpenClaw

​Step 2: Initialize Setup

​1. Confirm Installation

​2. Select Installation Mode

​3. Select Provider

​4. Select Models

​5. Select Default Model

​6. Select Channel

​7. Configure Skills

​8. Enable Hooks

​9. Restart Gateway Service

​10. Launch Bot

​Step 3: Configure EvoLink API & Model Switching

​1. Locate Two Configuration Files (Important)

​2. Configure Model Providers

​3. Configure Default Model (Supports Quick Switching)

​4. Quick Switch to EvoLink Models (Recommended)

​5. Restart and Verify

​Common Commands

​Troubleshooting

​Next Steps

Overview

System Environment Check (Optional)

Download the Check Tool

Check Items

Prerequisites

1. Install Node.js

2. Get EvoLink API Key

Step 1: Install OpenClaw

Step 2: Initialize Setup

1. Confirm Installation

2. Select Installation Mode

3. Select Provider

4. Select Models

5. Select Default Model

6. Select Channel

7. Configure Skills

8. Enable Hooks

9. Restart Gateway Service

10. Launch Bot

Step 3: Configure EvoLink API & Model Switching

1. Locate Two Configuration Files (Important)

2. Configure Model Providers

3. Configure Default Model (Supports Quick Switching)

4. Quick Switch to EvoLink Models (Recommended)

5. Restart and Verify

Common Commands

Troubleshooting

Next Steps