Skip to content

Personal AI agent

Tashi

Part of my work is supported by an agent. This is the setup behind it.

I’m Jodok Batlogg. Tashi helps me research, draft, organise and follow through. Some of my communication is prepared or handled with this support. I remain responsible for the work and the decisions.

Working together

Calm. Attentive. Precise.

Tashi is my personal AI agent, running on OpenClaw. A sparring partner for strategy, a second perspective on blind spots, and support for the routine work that still needs doing.

  • Speak little, speak clearly. No filler, no pathos.
  • Explain causes rather than assign blame.
  • Help without imposing. Earn trust through competence.
  • Make uncertainty explicit. Errors should never pass silently.

The name comes from Tibetan: good, auspicious, appropriate. The principles are Buddhist-inspired, grounded in practice.

OpenClaw setup

One agent. Connected tools.

OpenClaw connects the agent to tools, messaging and persistent memory. Coding work can be handed to Codex or Claude Code; scheduled tasks and webhooks bring work back to the agent.

The original setup ran on a Mac Mini M4 Pro in Dornbirn, Austria. The integrations listed here describe the recovered setup, rather than a live inventory.

OpenClaw source
  • Tools | Web search, browser automation, shell, files and GitHub.
  • Workspace | Gmail, Calendar, Google Workspace and HubSpot CRM.
  • Channels | WhatsApp, Telegram, iMessage and Slack.
  • Local access | 1Password and macOS automation with Peekaboo.
  • Voice | ElevenLabs and Whisper.
  • Continuity | Persistent files, vector memory search, heartbeats and scheduled tasks.

Where Tashi helps

From thinking to follow-through.

  • 01

    Strategic sparring

    Test assumptions, compare scenarios and find the question that matters.

  • 02

    Research and drafting

    Gather context, reduce complexity and prepare useful first drafts.

  • 03

    Communication

    Help prepare messages in my tone and language context.

  • 04

    Operations

    Support calendars, workflows, coding and the routine work between decisions.

Responsibility

Useful, with boundaries.

An agent can help me do more. It can also misunderstand context or get something wrong. If a message feels off, tell me. The responsibility stays with me.

  • Act when the task is clear. Ask when consequences are hard to reverse.
  • Private things stay private.
  • No half-finished outbound messages.
  • Persist context so the next task does not start from zero.

Feedback

Something went wrong?

If Tashi got something wrong or treated you unfairly, please send the context. You will receive a response by email.

Your submission is processed with AI support. See our privacy policy.

Contact Jodok directly