How to Deploy OpenClaw – Autonomous AI Agent Platform
OpenClaw (formerly Moltbot) is a personal AI assistant you can run on your own devices. It connects to popular messaging platforms like WhatsApp, Telegram, Slack, Discord, and many more, providing a unified assistant experience across channels. OpenClaw runs locally and includes a Gateway control plane that manages sessions, routes messages, executes tools, and maintains persistent memory.
This article explains how to deploy OpenClaw using Docker Compose with its interactive setup wizard. It covers model configuration, channel integration, gateway setup, the persistent memory system, and optional integration with Vultr Serverless Inference.
Prerequisites
Before you begin, you need to:
- Have access to an Ubuntu 24.04-based server as a non-root user with
sudoprivileges. - Install Docker and Docker Compose.
- Have an API key for a supported model provider. You can also use Vultr Serverless Inference, which we configure later.
- (Optional) Configure a DNS A record pointing to your server's IP address (for example,
openclaw.example.com).
Deploy OpenClaw Using Docker Compose
The OpenClaw repository includes a setup script that handles building, onboarding, and starting the gateway. For the complete Docker setup guide, see the official Docker documentation.
The interactive wizard configures model providers, channel integrations, and security settings in one streamlined flow.
Create the project directory and navigate into it.
CONSOLEClone the official OpenClaw repository and switch to the cloned directory.
CONSOLERun the Docker setup script.
CONSOLEThe interactive wizard walks you through:
- Security acknowledgment: Review security best practices for running AI agents.
- Model provider setup: Select your preferred model provider (Anthropic, OpenAI, Google, OpenRouter, and so on) and enter your API key or authenticate via OAuth.
- Channel configuration: Choose messaging platforms (Slack, Discord, Telegram, WhatsApp, and so on) and enter the required tokens.
- Channel allowlists: Configure which channels or users can interact with the bot.
- Skills setup: Enable optional capabilities like web search, image generation, and more.
- Gateway startup: Builds the image, generates a gateway token, and starts the service.
Verify the gateway is running by checking the container logs.
CONSOLEThe output shows a successful startup:
The wizard writes configuration and workspace data to:
~/.openclaw/: Configuration, credentials, and session data~/.openclaw/workspace: Agent workspace (memory files, skills, and so on)
Access the OpenClaw Control UI
The Control UI provides a web interface for managing your OpenClaw installation.
Note
For security, the Control UI requires either localhost access or HTTPS. It will not work over plain HTTP from a remote IP.
Retrieve your gateway token from the configuration file.
CONSOLEOutput:
Choose one of the following access methods:
Explore the main interface sections:
- Chat: Interact with the assistant directly through the built-in chat interface.
- Overview: View gateway status and health information.
- Channels: Monitor messaging platform connections and their status.
- Sessions: View active chat sessions across all connected channels.
- Skills: Browse available agent skills and capabilities.
- Config: Manage gateway and agent configuration settings.
Use the Chat interface to verify the installation works correctly.
Use Vultr Serverless Inference (Optional)
OpenClaw supports custom model providers through OpenAI-compatible endpoints. Configure Vultr Serverless Inference as a model provider to access additional AI models.
Note
OpenClaw requires a model that supports tool calling. Vultr model availability and tool-calling support can change over time, so verify the current supported models through the Inference API before selecting one for OpenClaw. At minimum, use a model that successfully returns tool_calls from the chat/completions endpoint during testing.
Enable Vultr Serverless Inference in the Vultr Console and copy your API key.
Edit the OpenClaw configuration file.
CONSOLEAdd the Vultr provider to the
modelssection. ReplaceYOUR-VULTR-API-KEYwith your Vultr API key.JSONSave and close the file.
Restart the gateway to apply the new provider.
CONSOLESelect the Vultr model in the Control UI or via the chat command.
INIVerify the model responds correctly by sending a test message.
OpenClaw Memory Database
OpenClaw maintains persistent memory through local file storage, enabling continuous learning across conversations. The system stores session transcripts, conversation history, and long-term memories to provide context-aware assistance that improves over time.
Session Storage
The Gateway stores all session data locally on the host.
Memory Files
OpenClaw maintains two layers of memory through Markdown files in the agent workspace.
- Daily Logs (
memory/YYYY-MM-DD.md): Append-only logs for day-to-day context. The assistant reads today's and yesterday's logs at the start of a session. - Long-term Memory (
MEMORY.md): Curated facts, preferences, and decisions that persist across sessions.
To instruct OpenClaw to remember something, send a message like:
The assistant writes this to the appropriate memory file for future reference.
Session Management Commands
Control OpenClaw sessions directly from any connected chat.
/status: View current session status (model, tokens, cost)/newor/reset: Start a fresh session/compact: Summarize and compress session context/think <level>: Adjust thinking depth (off, minimal, low, medium, high)
Backup and Migration
Back up your OpenClaw data.
CONSOLERestore on a new server.
CONSOLE
Use Cases
- Personal productivity assistant: Track tasks, set reminders, draft emails, and summarize documents directly from Slack, WhatsApp, or any connected channel.
- Multi-channel unified inbox: Start a conversation on one platform and continue on another; OpenClaw maintains context across all connected channels.
- Team collaboration bot: Deploy in Slack channels to answer questions, search documentation, and assist with workflows.
- Development and DevOps helper: Get coding assistance, run commands, and automate routine tasks via tool integrations.
- Scheduled automation: Use cron jobs, webhooks, and Gmail Pub/Sub to trigger assistant actions automatically.
- Persistent knowledge base: Build long-term memory that persists across sessions and improves over time.
Conclusion
You have successfully deployed OpenClaw as a personal AI assistant with Docker Compose. The interactive setup wizard streamlines the setup process by configuring your model provider, messaging channels, and gateway in one flow. The memory database ensures continuity across sessions by building a personalized knowledge base that improves over time. With Vultr Serverless Inference, you can access additional AI models like Kimi K2 directly on Vultr's infrastructure. For more information, refer to the official OpenClaw documentation.
0 Comments
Be the first to comment and share your perspective with the community.