12 Matching Annotations
  1. Last 7 days
    1. Clawdbot Realistic Costs:Software: Free (MIT licensed, forever)Hardware: VPS $4-5/month, or Raspberry Pi ~$50-100 upfront, or old laptop free, or Mac Mini ~$600AI Model: Claude Pro $20/month (casual) to Claude Max $200/month (heavy use like Viticci)Realistic minimum: ~$25/monthBut remember: that $300+ in 2 days user is real. Heavy agentic use burns through tokens fast.

      Assuming cloud based models. Why? You could drop up to 200 month on a VPS and be really self sufficient, but probably wouldn't need a VPS that heavy?

    1. Clawdbot is built for something completely different. Think of it as a personal executive assistant that lives inside your messaging apps. Just like you’d email instructions to a real assistant who then handles the work while you focus on other things, Clawdbot receives natural language requests through WhatsApp, Telegram, Slack, or Discord and then executes actual tasks on your computer.

      Description of Clawdbot/ now named openclaw, as a PA

    2. Clawdbot uses AI APIs on every interaction and every background task. Costs accumulate.

      need to understand better where it is getting its models from. Can be local too? Then costs are your own compute only. Why is everyone always assuming cloud models against fee?

    3. Clawdbot requires technical setup. You need to deploy it on a server or local machine, configure messaging platform integration, manage API keys, and set permission boundaries thoughtfully. It’s not a consumer app. You can’t open an app store, tap install, and have it ready. That’s a real friction point that’s worth acknowledging.

      run on a dedicted machine?

    4. Use Clawdbot when your primary bottleneck is the accumulation of small tasks across your digital life.

      clawdbot for 'small tasks across your digital life', vgl [[Aazai 2025 aantekeningen]] wrt such tasks.

    5. Clawdbot’s power to access messaging platforms means anyone with a security compromise at any layer could potentially impersonate you to the agent. A prompt injection through a web page it’s browsing, a malicious message in a group chat, or a crafted email could theoretically redirect it toward unintended actions. Proper sandboxing and permission boundaries mitigate this, but they require genuine technical discipline

      clawdbot as additional attack surface

    6. Clawdbot requires elevated system permissions to do what it does. It needs to read and write files, execute shell commands, access your terminal, connect to services on your behalf. Running an always-on agent with access to your credentials, your messaging platforms, and your file system creates security surface area that’s worth understanding.

      security concerns wrt clawdbot bc it executes actions you'd normally do.

    7. Every conversation you have with it, every preference you state, every decision you make gets stored in a markdown file that evolves over time. Future requests pull relevant context from that history automatically. You don’t have to remind it of things.

      Full context maintained in a md file.

    8. Clawdbot is built for async, long-running, continuous tasks that exist across your entire digital life. It’s the tool that monitors your inbox at midnight, processes information while you sleep, and sends you a briefing in the morning. It’s designed to be always on, always learning your preferences, always available through the messaging apps where you already live

      clawdbot is persistent across time / tasks.

    9. It maintains persistent memory across conversations, so it remembers your preferences, past decisions, and ongoing projects. It can monitor scheduled tasks, send you proactive notifications, and continuously work on long-running tasks even when you’re not actively messaging it.

      clawdbot maintains context over time In a .md log of sorts?

    10. Clawdbot connects to dozens of services by default: Gmail, Google Calendar, Todoist, GitHub, Spotify, even smart home devices. When it needs capabilities it doesn’t have built in, it can request them, and with proper guidance from you, it can expand those capabilities itself.

      'by default' or can do it out of the box, if switched on?