From one command to an AI-transformed company

1. One command builds the entire infrastructure for running AI agents securely

   Keys, containers, gateway, permission service, database and chat are all created automatically. Every minion is its own OpenClaw instance in its own container. Services run in separate networks, and the database sits in an internal network with no access from outside.

2. Connected to AD or SSO, files, email, databases and any AI model

   Active Directory, Entra ID, Keycloak, Microsoft 365, SMB, Exchange, databases, REST and many other systems are ready to go. You choose the models freely: Bedrock, Azure, Google, OpenAI, Anthropic, a local model or any OpenAI-compatible server.

3. One minion administers the entire system, right in the chat

   What would otherwise mean clicking through the admin area, the System Minion does on request: creating minions, connecting and testing data sources, granting approvals. For new data sources it builds the right tools itself.

4. A Minion DB becomes the central interface where agents read and write

   Your data comes together in one database: PostgreSQL with full-text and vector search, plus originals and versions. Agents work through this one interface, reading sources and writing results, all locally on your server.

5. Data protection impact assessment, works council documents and AI training are generated

   The System Minion writes the draft impact assessment from your installation details and updates it when connectors, departments or model providers change. Records of processing, deletion concept, technical measures, employee information and works council documents come as templates in the package.

6. Security: own containers, 2FA, audit log, access only for approved tasks

   Every minion runs in its own container and never sees the model providers' keys. Admins sign in with a second factor, tools that need approval run only in approved tasks, and the audit log records what happens.

7. Minions and the right to use them come from AD

   Your AD groups automatically become minions, and the members of a group can use their minion. When a group changes, the system updates the assignment at the next sync. Without AD you set this up in the interface.

8. Employees set up and complete the AI training required by the AI Act

   Once per person: details about their work, email and files, task ideas and a training with a quiz. Passing is stored as proof.

9. Minions find the best task ideas themselves: no more blank white box

   The minion searches files, emails and calendars and comes up with ideas no human would ever have had. That solves the problem of the blank white box: two prompts are enough to AI-transform a department.

10. Ideas become script tasks, AI-checked and approved in the background

    All of this happens in the background. The end user notices none of it and simply chats with their minion. The minion builds the task and names the tools and data areas it needs. An AI checks this against your impact assessment, then you approve, or within known permissions it happens automatically.

11. Every task gets a task ID: only its tools and data, revocable at any time

    Tool permissions are calculated live from approved tasks; there is no second copy. That is why a revocation takes effect on the very next call.

12. Minions can do almost anything. Anything can become a task

    Minions take over entire processes and automate them from start to finish. Tasks start on a schedule, on new messages and emails, or on request, and results appear as a table, checklist, chart, metric, form or kanban board.

13. Minions optimize themselves, including on cost

    Every minion regularly reviews its work and looks for concrete improvements to tasks, scripts and collaboration. Cost too: for simple work it switches itself to a dumber, cheaper model, and for hard work to a smarter one.

14. Task by task, more runs by itself, all the way to an AI-transformed company

    More and more tedious work gets automated. And people get time for people again.