Skip to content

Documentation

Everything you need to deploy, configure, and manage your AI agent.

Getting Started

OCLauncher deploys a personal AI agent in a secure cloud sandbox. Your agent runs 24/7 on Cloudflare's global edge and connects to messaging platforms like Telegram and WhatsApp.

1. Create an account

Sign up with email or Google OAuth. It takes a few seconds.

2. Subscribe

Managed OpenClaw hosting is US$19/month. Get a dedicated instance and connect all your channels.

3. Deploy

Your AI agent deploys in under 30 seconds. Once running, it's always on and accessible from any connected channel.

Dashboard

The dashboard is your central hub for managing your AI agent instance.

The main panel shows your instance status, connected channels, and provider configuration. From here you can restart, update, or destroy your instance.

The Providers page lets you manage API keys for AI model providers (Anthropic, OpenAI, Google, OpenRouter).

The Channels page shows all connected messaging platforms and lets you add or remove connections.

Quick actions include restarting your instance, opening the web chat, and accessing the in-browser terminal.

Messaging Channels

Connect your AI agent to messaging platforms so you can interact with it from anywhere.

Telegram

Create a bot via @BotFather, copy the token, and paste it in the Channels page. Your agent starts receiving messages immediately.

Supports text messages, images, and group mentions. Configure DM access policies (allowlist or open).

WhatsApp

Pair using the QR code on the Channels page. Your agent appears as a linked device on your WhatsApp account.

Supports text and media messages. Configure self-chat mode and DM access policies.

Discord

Create a bot at discord.com/developers, enable Message Content Intent, and paste the token. Invite the bot to your server.

Supports server messages, DMs, and thread replies. Configure DM policies (pairing code, allowlist, or open).

Slack

Create an app at api.slack.com, generate App-Level and Bot tokens, then paste both in the Channels page.

Supports workspace messages and DMs. Requires connections:write and chat:write scopes.

More Platforms

Signal, Google Chat, Microsoft Teams, LINE, Matrix, and Feishu are available via CLI configuration inside the terminal.

AI Providers

Choose how your AI agent accesses language models.

Platform Key

Every plan includes a Google Gemini key managed by OCLauncher. No configuration needed — it works out of the box.

Bring Your Own Key (BYOK)

Add your own API keys from Anthropic, OpenAI, Google, or OpenRouter. This unlocks additional models like Claude, GPT-4, and more.

Key Security

All API keys are encrypted with AES-256-GCM before storage. Keys are only decrypted when configuring your instance.

Skills

Skills extend your AI agent with new capabilities — no coding required.

Browsing Skills

Open the Skills Hub from your dashboard sidebar. Search and filter community-built skills by category.

Installing Skills

Click Install on any skill card. Your agent will be reconfigured automatically to use the new skill.

Web Chat Widget

Add a floating chat bubble to any website. Visitors click it, talk to your AI agent in real time. Works on WordPress, Shopify, Squarespace, static HTML, React apps — anything that supports a script tag. No coding required.

Getting Started

1. Activate the Widget

Go to Dashboard > Plugins, find Web Chat, and click Activate. This generates a unique widget token and deploys the chat bridge to your instance. Takes about 3 seconds.

2. Customize

Switch to the Customize tab to configure theme (light, dark, or custom colors), primary color, greeting message, agent name, which agent handles chats, system prompt, and widget position (bottom-right or bottom-left). The live preview shows exactly how the widget will look.

3. Add to Your Website

Copy the embed code from the Overview or Settings tab and paste it just before the closing body tag on any page where you want the chat bubble.

Embed code:

<script src="https://oclauncher.com/widget/chat.js?t=YOUR_TOKEN"></script>

Platform-Specific Instructions

  • WordPress — Appearance > Theme File Editor > footer.php, or use a Header/Footer Scripts plugin
  • Shopify — Online Store > Themes > Edit Code > theme.liquid, before the closing body tag
  • Squarespace — Settings > Advanced > Code Injection > Footer
  • Wix — Settings > Custom Code > Body - End
  • React / Next.js — In your root layout component, add a Script tag or insert in _document.tsx
  • Webflow — Project Settings > Custom Code > Footer Code

4. Test It

Open your website in a browser. You should see the chat bubble in the corner. Click it, send a message, and verify you get a response from your agent.

Dashboard Features

Overview Tab — See total conversations, widget status (green pulse = live), your widget token, and a quick embed code section with one-click copy.

Conversations Tab — Browse all conversations the widget has handled. See role indicators (green = agent spoke last, gray = visitor), relative timestamps, message counts. Click any conversation to view the full message history.

Customize Tab — Three control groups: Appearance (theme, colors, position), Agent & Behavior (agent selection, system prompt), and Content (greeting, display name). The Preview panel updates live as you change settings.

Settings Tab — Widget status indicator, full embed code with terminal-style display, usage instructions, and a danger zone to deactivate the widget.

Customization

Light theme works best for light-colored websites. Dark theme matches the OCLauncher dashboard aesthetic. Custom theme gives you full control over panel background, message bubble color, text color, and border color.

Good system prompts are concise and specific. Tell the agent who it is, what it should help with, and what to avoid. Set clear boundaries and include escalation paths (e.g., 'For complex issues, suggest they email support@example.com').

If you have multiple agents configured, you can route widget conversations to a specific one — useful for a dedicated support agent or a different personality for web visitors. Leave it as Default to use your primary agent.

How It Works

When a page loads, the widget script fetches your customization settings. The chat bubble appears in the configured corner. Visitors click it to open the panel with your greeting. Messages are sent to your AI agent in real time, and responses stream back word-by-word. Conversations persist for the browser session and are cleared when the tab closes.

Conversation data is stored on your Fly.io instance, not on OCLauncher's servers. The widget uses sessionStorage only (no cookies), loads no third-party tracking scripts, and renders inside a Shadow DOM for complete CSS isolation from your site.

The widget script is ~15KB (~5KB gzipped), has zero external dependencies, and renders in under 100ms. Shadow DOM ensures no CSS conflicts or layout shifts on your site.

Troubleshooting

Widget doesn't appear — Check that the token in the script tag is correct. Look for [OCL Chat] warnings in the browser console. If your site has a Content-Security-Policy, add oclauncher.com to script-src. Verify your Fly instance is running.

Messages fail — Check your instance status on Dashboard > Overview. Verify the Web Chat plugin is active under Plugins. If you see 'Please wait a moment,' you've hit the per-IP rate limit — wait 60 seconds. The system auto-repairs the chat bridge if it detects issues.

Styling conflicts — The widget uses Shadow DOM to prevent CSS conflicts. If something covers it, check for z-index conflicts (widget uses max value). If your site has a fixed footer in the same corner, switch the widget to the opposite position.

Limits

10 messages per minute per visitor IP. 60 messages per minute per widget token. Last 20 messages sent as context to the agent. File-based conversation storage on your instance (scales to ~10k conversations). One widget token per account. Script size ~15KB (~5KB gzipped).

Terminal

Access a terminal session inside your sandbox container directly from the browser.

Accessing the Terminal

Open the Terminal page from the dashboard sidebar. The session connects to your running sandbox via WebSocket.

Use Cases

Run OpenClaw CLI commands, inspect logs, configure additional channels, manage files, and debug your agent's behavior.

Storage & Data

Each instance includes persistent storage that survives restarts and updates.

Capacity

8 GB of persistent storage per instance, backed by Cloudflare R2. Files, agent memories, and configurations are all preserved.

Persistence

Data persists across instance restarts and updates. Storage is isolated per user — no other user can access your files.

Lifecycle

Storage is provisioned when your instance is created and deleted when the instance is destroyed. Cancelling your subscription will eventually destroy the instance.

Security

OCLauncher is built with security-first principles at every layer.

Sandbox Isolation

Every user gets a fully isolated sandbox container. Containers cannot access each other's data, network, or processes.

Encryption

All connections use TLS. API keys are encrypted at rest with AES-256-GCM. Passwords are stored as cryptographic hashes.

Infrastructure

Hosted on Cloudflare's global edge network with built-in DDoS protection, WAF, and automatic TLS certificate management.

Billing

Simple, transparent pricing with no hidden fees.

Pricing

Managed OpenClaw hosting is US$19/month. It includes a dedicated instance, all messaging channels, and platform AI key.

Getting Started

Subscribe and your instance is ready in 30 seconds. 3M tokens are included from your first billing cycle.

Cancellation

Cancel anytime from the Billing page. Your instance stays active until the end of the current billing period.

Money-Back Guarantee

New subscribers get a 7-day money-back guarantee. Email support@oclauncher.com within 7 days for a full refund, no questions asked.

Tokens

AI model usage is measured in tokens. Your plan grant refreshes each billing cycle. If you run out, you can purchase additional token packs.

Token Packs

Buy extra tokens in one-time packs from the Billing page. Purchased tokens don't expire and carry over between billing cycles.

Mini Apps

Mini apps are interactive single-page tools built by your AI agent. They persist data, match the platform theme, and are accessible from your dashboard.

What Are Mini Apps?

When you ask your agent to build a tool — a tracker, dashboard, form, timer, or any interactive app — it creates a mini app. Each mini app is a self-contained web page with persistent key-value storage, themed to match the platform. You can access them from the Mini Apps page on your dashboard.

Creating a Mini App

Just ask your agent in any channel: "Build me an expense tracker" or "Create a kanban board for my tasks." The agent scaffolds the app, writes the HTML, and presents it to you. No coding required on your part.

Managing Mini Apps

The Mini Apps page on your dashboard shows all your apps as cards. You can edit the title, description, and icon inline. Pin favorites to keep them at the top. Open apps in an embedded viewer or fullscreen mode, or launch them in a new browser tab.

Persistent Storage

Each mini app can store up to 100 keys, with each value up to 1 MB. Data persists across page refreshes and instance restarts. Use the storage inspector on any app card to view stored keys and their values.

Example Use Cases

Expense trackers, project task boards, habit trackers, quiz apps, countdown timers, data dashboards, client feedback collectors, daily journals, inventory managers, and recipe collections. If it fits on a single page and needs persistent data, it's a good candidate for a mini app.

Tips

Pin frequently-used apps so they always appear at the top. Use the storage inspector to see what data an app is holding. Ask your agent to add toast notifications for user feedback, loading spinners for async operations, or confirmation dialogs before destructive actions.