OpenClaw

OpenClaw is an open-source AI agent platform that can connect to any OpenAI-compatible inference provider. This guide walks through configuring OpenClaw to use CompactifAI as its reasoning model via the OpenAI Chat Completions API.

Before you begin, ensure you have:

• A CompactifAI API key
• The model ID you want to use (for example carina-60b, carina-60b, or glm-5-2)
• Your CompactifAI API endpoint (typically https://api.compactif.ai/v1)

How it works

OpenClaw  -->  CompactifAI API

OpenClaw sends requests directly to CompactifAI using the OpenAI-compatible /v1/chat/completions endpoint. No proxy or translation layer is required.

1. Install OpenClaw

macOS / Linux

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

Windows (PowerShell)

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

Alternatively, if you already have Node.js installed:

npm install -g openclaw@latest

Danger

[IMPORTANT] Press Ctrl + C or type exit and Exit the installation at this point.

2. Start the classic onboarding wizard

Launch the interactive setup:

openclaw onboard --classic

When prompted:

• Accept the security disclaimer.
• Select Manual setup.
• Choose Local gateway (this machine).
• Accept the default workspace directory unless you have a reason to change it.

3. Configure CompactifAI

When prompted for Model/auth provider, select:

More…
└── Custom Provider

Then enter the following values:

Prompt	Value
API Base URL	https://api.compactif.ai/v1
API Key	Your CompactifAI API key
Endpoint compatibility	OpenAI-compatible
Model ID	A model ID from the Models Catalog (for example carina-60b, carina-60b, or glm-5-2)
Endpoint ID	Accept the generated value (custom-api-compactif-ai) or enter your own
Model alias	Optional
Does this model support image input?	Select Yes for multimodal models; otherwise No

OpenClaw verifies the endpoint by sending a test completion request. Continue once you see:

Verification successful.

4. Configure the Gateway

The onboarding wizard then configures the local gateway.

Unless you have specific networking requirements, accept the default values:

Prompt	Recommended value
Gateway port	18789
Gateway bind address	Loopback (127.0.0.1)
Gateway access protection	Token (recommended)
Tailscale exposure	Off
Gateway token	Generate a new token or reuse an existing one

5. Configure Chat Channels

OpenClaw supports connecting to messaging platforms including Discord, Slack, Telegram, WhatsApp, Signal, Microsoft Teams, Matrix, and many others.

When prompted:

Set up a chat channel now?

Choose:

• Yes to configure a messaging platform immediately.
• No to finish the installation and configure channels later.

You can add or update channels at any time after installation.

6. Launch OpenClaw

Start OpenClaw:

openclaw

or launch the default assistant directly:

openclaw crestodian

OpenClaw will use the CompactifAI model you configured during onboarding.

7. Verify the Connection

Try a simple prompt:

Hello! Which model are you using?

or

Write a Python function to calculate Fibonacci numbers.

If the model responds successfully, OpenClaw is communicating with CompactifAI correctly.

Requests are sent using the same OpenAI-compatible Chat Completions API described in Chat Completion.

Troubleshooting

Issue	Solution
401 Unauthorized	Verify your CompactifAI API key from the Authentication page.
404 Model Not Found	Ensure the model ID exactly matches a deployed model from the Models Catalog.
Verification failed	Confirm the API Base URL is correct and that the endpoint implements POST /v1/chat/completions.
Connection failed	Verify the endpoint is reachable and includes the /v1 suffix.
Streaming errors	Ensure your deployment supports Server-Sent Events (SSE) for streaming responses.
Gateway unavailable	Verify that the local gateway is listening on 127.0.0.1:18789, or rerun openclaw onboard --classic.

For details about request and response formats, see Chat Completion and the API Reference.