Getting Started

clawd.run gives you a persistent AI agent that lives across your messaging apps. It remembers conversations, runs code, searches the web, and connects to the channels you already use.

Create an account

Chat in the dashboard

Start a conversation directly in the browser. Web search, code execution, and memory work from day one.

Connect a channel

Plug in Telegram, Discord, Slack, or any of 17 channels. Your agent responds like a native bot.

Customize

Edit your agent's personality, switch models, bring your own API keys, and configure tools in Settings.

Channels

Connect your agent to the messaging platforms you already use. Each channel runs the same agent. Conversations are unified across all of them.

Invite-enabled accounts can connect Telegram via @BotFather and a bot HTTP API token.

Required: Bot Token (from @BotFather)

Discord

Invite-enabled accounts can connect a Discord bot token from discord.com/developers.

Required: Bot Token (from Discord Developer Portal)

Slack

Invite-enabled accounts can connect Slack using Bot Token and App-Level Token.

Required: xoxb- Bot Token + xapp- App Token

Invite-enabled accounts can connect by scanning a QR code with WhatsApp. No developer account needed.

Scan QR code in Dashboard - Channels - WhatsApp

17 channels including Matrix, IRC, Nostr, Mattermost, LINE, Twitch, and more. Token and QR channel access is controlled through invites.

Memory

Your agent maintains memory across every conversation, channel, and session. Nothing resets between chats.

Conversation memory

The agent remembers what you discussed, preferences you've stated, and tasks you've assigned — automatically, across all your connected channels.

Long-term facts

Important information is extracted and stored in a persistent knowledge base. Tell the agent your timezone, preferences, or project context once. It remembers.

File memory

Your agent has a persistent workspace. Files you create, upload, or generate stay available across sessions.

Memory search

Ask the agent to recall anything: 'What did we decide about the landing page?' — it searches its full history.

Tools

Your agent ships with a full toolkit. No configuration needed — tools are available from the first message.

Web Search

Real-time search with source citations. Stays current beyond training data cutoff.

Code Execution

Write and run Python, Node.js, or bash. Results stream back into the conversation.

File Management

Create, read, edit, and organize files in your persistent workspace.

Image Generation

Generate images from text prompts using state-of-the-art models.

Browser Automation

Navigate websites, fill forms, and extract data from any URL.

Scheduling

Set reminders and recurring tasks. The agent proactively reaches out.

Bring Your Own Keys

Connect your own API keys to use your existing credits, access private models, or bypass rate limits.

Your keys are stored encrypted in your isolated sandbox. They are never shared with other users or used for any other purpose.

Anthropic

Claude Opus, Sonnet, and Haiku. Go to console.anthropic.com to get your key.

OpenAI

GPT models and image generation. Go to platform.openai.com to get your key.

xAI

Grok models. Go to console.x.ai.

Google

Gemini models. Go to aistudio.google.com to get your key.

Fal.ai

Image and video generation. Go to fal.ai/dashboard to get your key.

Add keys in Dashboard - Settings - API Keys.

Model Selection

Model availability depends on your infrastructure profile. Switch models mid-conversation when available.

BaseMiniMax, Sonnet, Grok FastDefault routing

ExtendedAdds GPT and Gemini familiesInvite-enabled

FullAll available models including premium reasoningManual enablement

Configuration

Your agent is configured via a JSON file in Dashboard - Config. Changes apply instantly.

{
  "agent": {
    "name": "My Agent",
    "model": "claude-sonnet-4",
    "personality": "Direct and concise."
  },
  "tools": {
    "web_search": true,
    "code_execution": true,
    "image_generation": true
  },
  "memory": {
    "retention": "forever"
  }
}

agent.name

The name your agent introduces itself with in channels.

agent.personality

A short system prompt that defines tone and behavior. Keep it under 500 characters.

tools

Toggle individual tools on or off. Useful for scoping what the agent can do.

FAQ

Is my data private?

Yes. Your conversations, files, and config are stored in an isolated environment. We do not train on your data and do not share it with other users or third parties.

Can I export my data?

Yes. Go to Dashboard and download your conversation history, files, and config at any time.

How does access work?

clawd.run currently uses invite-based access. Request access on the homepage, then sign in once your email is approved.

Can my agent message me proactively?

Yes. Set up recurring tasks or one-time reminders in chat. The agent will reach out on your connected channels at the specified time.

How are models selected?

Model routing is configured per account. The platform can auto-route, or you can pin a model from the dashboard when available.

What are credits?

Credits measure the cost of each AI response. Grok 4.1 Fast = 1 credit. Claude Sonnet = 3 credits. Claude Opus = 10 credits. Your pool resets on your billing date each month.

Still have questions?

support@clawd.run