# Run AI CoWork Locally with OpenClaw & Ollama (Complete Setup)

Source: https://www.youtube.com/watch?v=V6Mp0fUh-OM
Recap page: https://rapidrecap.app/video/V6Mp0fUh-OM
Generated: 2026-02-01T08:33:26.675+00:00

---
## Quick Overview

The video provides a complete setup guide for running an AI coworker locally using OpenClaw, which acts as a gateway, integrated with Ollama for model execution, allowing users to control their computer via messaging apps like Slack while keeping data private.

**Key Points:**
- The AI assistant, originally named Clawdbot, has been officially renamed to OpenClaw.
- OpenClaw installation requires running `npm install -g openclaw@latest` followed by `openclaw onboard --install-daemon`.
- OpenClaw integrates with Ollama to run models locally, ensuring user data remains 100% private.
- The video demonstrates setting up Slack integration by configuring OAuth & Permissions, enabling Socket Mode, and subscribing to necessary events in the Slack API Console.
- Ollama must be installed first, and models like `gpt-oss:20b` are downloaded locally using `ollama pull model_name`.
- The configuration process for web UI access involves running `openclaw configure --section web` and selecting local gateway execution.
- The final result demonstrates controlling the local computer (e.g., listing directory contents) via commands sent through the OpenClaw web UI or directly via Slack.

![Screenshot at 00:02: The screen displays the transition from the old name 'Clawdbot' to the new official name 'OpenClaw' on the landing page, emphasizing its core function as 'THE AI THAT ACTUALLY DOES THINGS.'](https://ss.rapidrecap.app/screens/V6Mp0fUh-OM/00-00-02.jpg)

**Context:** This tutorial details the process of setting up OpenClaw, an AI personal assistant that runs locally on the user's device, to manage tasks and bridge various messaging services (like WhatsApp, Telegram, and Slack) to AI coding agents via a centralized gateway. The system utilizes Ollama to run large language models (LLMs) locally, emphasizing data privacy by keeping all processing on the user's machine rather than in the cloud. The video walks through installation, Ollama setup, and the specific configuration steps required for Slack integration.

## Detailed Analysis

The video guides the viewer through setting up OpenClaw, an AI assistant that integrates with messaging apps and runs local LLMs via Ollama. The presenter first explains the naming journey, noting the project rebranded from Clawdbot to OpenClaw. Installation starts by installing OpenClaw globally via npm: `npm install -g openclaw@latest`, followed by running the onboarding wizard: `openclaw onboard --install-daemon`. A key configuration step involves setting the context length in Ollama settings to at least 64k tokens for better performance. Next, the presenter details integrating with Ollama by running `ollama launch openclaw` (noting that the old alias `ollama launch clawdbot` is deprecated). After downloading a model locally using `ollama pull gpt-oss:20b`, the user accesses the OpenClaw web UI (http://127.0.0.1:18789/). The video then transitions to setting up Slack integration using the documentation provided by Praison AI. This involves creating a Slack app, configuring OAuth & Permissions with specific scopes (`chat:write`, `app_mentions:read`, `im:history`, `im:read`, `channels:history`), enabling Socket Mode with `connections:write` scope, subscribing to necessary events (`app_mention`, `message.im`), and finally reinstalling the app. The demonstration concludes by showing that commands sent via the Slack interface (like listing files in a directory) are successfully executed by the locally running OpenClaw gateway.

### Naming Journey & Installation

- Clawdbot renamed to OpenClaw
- Initial installation via npm: `npm install -g openclaw@latest`
- Run onboarding wizard: `openclaw onboard --install-daemon`
- Set Ollama context length to 64k tokens in settings.

### Ollama Setup

- Use `ollama pull gpt-oss:20b` to download models locally
- Launch OpenClaw integration with `ollama launch openclaw` (old alias `clawdbot` deprecated).

### OpenClaw Web UI Configuration

- Run `openclaw configure --section web` in the terminal
- Select 'Local (this machine)' for Gateway execution
- Access Web UI at http://127.0.0.1:18789/.

### Slack Integration Steps

- Create Slack App via Slack API Console
- Configure OAuth & Permissions with required scopes (e.g., `chat:write`, `im:read`)
- Enable Socket Mode with `connections:write` scope
- Subscribe to bot events (`app_mention`, `message.im`)
- Reinstall the app after configuration changes.

### Demonstration

- Successfully query the bot via Slack to list directories in a local folder (`List files in my ~/praisonai-package folder`) to confirm local execution and control.

![Screenshot at 00:02: The transition screen showing Clawdbot rebranding to OpenClaw, highlighting its claim as 'THE AI THAT ACTUALLY DOES THINGS.'](https://ss.rapidrecap.app/screens/V6Mp0fUh-OM/00-00-02.jpg)
![Screenshot at 00:02: The installation documentation showing the required npm command: npm install -g openclaw@latest and the subsequent onboarding command.](https://ss.rapidrecap.app/screens/V6Mp0fUh-OM/00-00-02.jpg)
![Screenshot at 00:20: Ollama model repository listing available models like kimi-k2.5 and glm-4.7-flash, which can be downloaded locally.](https://ss.rapidrecap.app/screens/V6Mp0fUh-OM/00-00-20.jpg)
![Screenshot at 02:42: The Ollama settings window showing the context length slider being set to 64k tokens, as recommended for OpenClaw.](https://ss.rapidrecap.app/screens/V6Mp0fUh-OM/00-02-42.jpg)
![Screenshot at 03:30: Terminal output confirming that the ollama launch clawdbot command successfully configured the gateway with Ollama, despite initial errors indicating the gateway was already running.](https://ss.rapidrecap.app/screens/V6Mp0fUh-OM/00-03-30.jpg)