First-time user tip: If you haven’t deployed your own OpenClaw yet, we recommend getting started with Tencent Cloud OpenClaw — deploy in just a few seconds with one click.
👉 Complete Tutorial Guide
This guide answers the most frequently asked questions during setup and usage, organized into the following sections:
Conversation Tuning
| 🎮 General FAQs | |
| I’m not sure how to install it. Will there be any offline installation events or meetups? | A series of offline events is being prepared—stay tuned by following the Tencent Cloud Server official account for the latest updates. We also host weekly online livestreams with step-by-step installation guidance. You can join the OpenClaw Cloud Community; or join the Lighthouse Official Community. |
| 🎨 Prerequisites & Purchase | |
| I want to deploy OpenClaw. What costs should I expect? | You’ll mainly need to purchase cloud server resources and LLM resources. Lighthouse (Lightweight Application Server) purchase: 👉 Click here LLM resources can be purchased from different providers based on your needs, and pricing varies. |
| There are many cloud server configurations. Which one should I choose? | Plan types: Entry (best value), High-Performance, General-Purpose. Plan specs: 2 vCPUs + 4 GB RAM or higher is recommended. The minimum supported spec is 2 vCPUs + 2 GB RAM (for basic trials). |
| Should I choose an overseas or Mainland China region for my cloud server? | Please choose the region based on your scenario and the models you plan to use. 1) Messaging apps: For QQ and WeCom, Mainland China regions are typically recommended; for Discord and WhatsApp, overseas regions are usually preferred. 2) Model access: If you plan to call Mainland China–hosted models, a Mainland China region is generally recommended. 3) Website access: If you want OpenClaw to retrieve information from international services such as GitHub and Notion, an overseas region is recommended. |
| My Tencent Cloud account registration keeps failing. Who can I contact? | You’ll need to complete identity verification for your Tencent Cloud account before purchasing. See the documentation, or submit a ticket on the official website for after-sales support. |
| After purchase, I can’t find the Lighthouse entry. Where do I configure OpenClaw? | Go to the Lighthouse Console. |
| 🤖 Deployment & Installation | |
| What’s the difference between local deployment and cloud deployment? | Local and cloud deployments each have their own characteristics. For users who want to get started quickly, we recommend cloud deployment, which offers: 1) Deploy in seconds: Stable public access with always-on service 2) Always online: Runs stably 24/7 with multi-device access 3) Security isolation: A dedicated cloud environment isolated from your local computer to better protect privacy, with professional security protections |
| My computer isn’t a Mac. Can I still deploy and install it? | Tencent Cloud Lighthouse deploys OpenClaw in the cloud, so there are no requirements for your computer’s performance or model—any computer can be used. |
| What does the deployment process look like? | Lighthouse provides an OpenClaw application template and a visual configuration panel—no command-line operations are required. The steps are roughly: 1) Purchase a cloud server and select the OpenClaw application template 2) Configure the LLM 3) Configure channels: WhatsApp, Telegram, Discord, Slack, QQ, Feishu, DingTalk, and WeCom |
| There are too many tutorials. Which one should I read first? | We recommend starting with the One-Click Deployment tutorial. The following resources are usually sufficient: 📚 Detailed Guides: One-Click Deployment: hhttps://www.tencentcloud.com/techpedia/139184 Complete Best Practices Collection: https://www.tencentcloud.com/techpedia/139189 |
| Why does my panel look different from what’s shown in the promo materials? | Please ensure you’re using the latest OpenClaw version. If you’re not on the latest version, go to Application Management on your server and click One-Click Update. Note: One-click update will cause channel configurations (Channels) except QQ to be cleared—you’ll need to reconfigure them after the update. Models, Skills, and Memory will not be lost. |
| ⚡️ Model Configuration | |
| Any other cost-effective model recommendations? | We recommend trying each vendor’s Coding Plan offerings first—they typically provide the best value by charging per call rather than by token usage. The community also compiled a guide: 2026 Free LLM API Quota Summary |
| Any tips to reduce token usage? | In chat apps, you can also use slash commands to quickly “shrink” conversations. Common commands:/compact Manually compress—summarize older content into a short digest/reset Keep memory but reset the current topic/new Start a brand-new conversationMore tips: Reduce Token Usage—Immediate Results |
| How do I get an LLM API Key? | In the Lighthouse model configuration panel, you can select Tencent Cloud Coding Plan, Tencent Hunyuan, or Tencent Cloud DeepSeek for one-click setup and activation. If you choose other model providers, you’ll need to obtain the API key from the provider’s official website. The configuration panel also provides guidance when you select a vendor. |
| 🔌 Channel Configuration | |
| Which messaging apps are currently supported? | Tencent Cloud Lighthouse initially supports integration with the four major IM tools in Mainland China: QQ, Feishu, DingTalk, and WeCom. |
| How do I connect to a messaging app? |
WhatsApp: https://www.tencentcloud.com/techpedia/139186 Telegram: https://www.tencentcloud.com/techpedia/139185 Discord: https://www.tencentcloud.com/techpedia/139187 Slack: https://www.tencentcloud.com/techpedia/139193 iMessage: https://www.tencentcloud.com/techpedia/139192 |
| I saw promo materials saying QQ supports installing 5 bots. How can I install multiple QQ bots? | Currently, one cloud server supports one bot. To run multiple bots, you’ll need multiple cloud servers. |
| The QQ bot replies: “This bot has gone to Mars. Please try again later.” | Possible causes: 1) You didn’t add the QQ bot’s App ID and App Secret to OpenClaw. 2) You added an incorrect QQ bot App ID/App Secret. Please follow the corresponding tutorial steps. If you’re using the latest configuration panel, this issue can usually be avoided. |
| After completing all tutorial steps, the QQ bot doesn’t reply in chat | Possible cause: Your model configuration may be incorrect, so the QQ bot cannot return responses. In the Lighthouse Console, go to Application Management and configure the model API Key. See: Deploy OpenClaw (formerly Clawdbot) on the Cloud in Seconds - Configure model API Key. How to verify model configuration: If you want to confirm the model setup is effective, log in to the Lighthouse instance and validate via the OpenClaw TUI tool. See: Deploy OpenClaw (formerly Clawdbot) on the Cloud in Seconds - Interact with OpenClaw using TUI. |
| Why can’t I connect to my company’s WeCom? | Connecting to WeCom requires admin privileges for the company’s WeCom tenant. |
| 🧩 Skills | |
| What are Skills? | If you think of the LLM as the “brain,” Skills are its hands and feet. What OpenClaw can ultimately do depends heavily on which Skills are installed. |
| Which Skills are installed by default in Lighthouse? | tavily-search Web searchsummarize Summarize a URL or fileagent-browser Simulate browser behaviorfind-skills Discover and install Skillsgithub Interact with GitHubobsidian Sync Obsidiannotion Manage pages, databases, and blocksweather Get weather and forecaststencentcloud-lighthouse-skill Manage Tencent Cloud Lighthousetencent-docs Tencent Docs capabilitiesIf you don’t use related Skills, no additional model tokens will be consumed. |
| How do I install a Skill? | Method 1: Install from the panel Browse, search, and install with one click in the admin panel Method 2: Install via chat Tell OpenClaw what Skill you need in chat—it will search and install automatically Method 3: Upload files and install manually Download the Skill package (ZIP), upload it via file management, then ask OpenClaw to install that Skill |
| Why do I get an error when I enter a Skill name in the configuration panel? | Please confirm the Skill name is correct. On SkillHub, click the Skill title; the short English identifier shown under the title is the Skill name. For example, the Skill name for “RedNote - Xiaohongshu” is xhs. |
| 💬 Conversation Tuning | |
| 401 Unauthorized Authentication Failed |
Cause 1: The model API Key itself is invalid • Incorrect API Key format 👉 A typical key looks like sk-xxxx…xxxx. Check for copy issues (line breaks, extra spaces, truncation, etc.) • API Key expired/invalid/deleted 👉 Regenerate the key or confirm its status • Mainland China vs International site mismatch 👉 Some vendors have different sites—ensure you selected the correct one Cause 2: The API Key does not match the selected model • Mixing API keys across platforms/models 👉 e.g., using a Kimi key while selecting MiniMax • Mixing API keys for the same model across different platforms 👉 e.g., a Tencent Cloud DeepSeek key cannot be used for the DeepSeek (DeepSeek Official) option • Mixing API keys across subscription types 👉 Coding Plan usually has a dedicated option and cannot be used for non–Coding Plan options Cause 3: Your model account is in an abnormal state • Account restricted/frozen/disabled • Account overdue or quota exhausted • Required verification/authorization not completed |
| 403 Forbidden Insufficient Permissions: This API Key is not authorized to call the model service |
• Check whether the model is supported in your server region (e.g., some overseas models may not be available in Mainland China regions) • Check whether your resource package/quota has been exhausted |
| 404 Not Found Incorrect Request: API path or resource ID does not exist |
• Check whether the model base_url is correct• Check whether the model ID is correct |
| 429/500 Insufficient Balance Usually indicates insufficient model account balance |
• Check whether the API Key is correct • Check whether the account is overdue |
| Rate_Limit Rate limit Usually indicates request rate limits from the model provider |
• Request a higher rate limit from the model provider • Switch to another model |
| OpenClaw isn’t replying. How do I troubleshoot? | TUI + logs — a 6-step troubleshooting flow 1) Log in to the server: access the command line of the server running OpenClaw 2) Enter TUI: run openclaw tui to open the interactive terminal UI3) Tail logs in another session: run openclaw logs --follow to view real-time logs4) Send messages and observe: send messages in both the chat app and TUI, and check whether logs look normal 5) Locate errors: search for keywords like Failed, Error, Not Found to pinpoint problematic lines 6) Analyze the cause: combine log details to identify likely causes, and use other AI tools to assist if needed |