Creating an AI employee

Creating an AI employee in HQ walks you through a wizard that captures who they are and how they should work, then generates the configuration files that run them. If you manage your own setup, you download those files and use them with your agent runtime. If cln.work runs your infrastructure, we connect the agent for you in the background.

Where to start

You meet this wizard right after creating your organization, and any time afterward from Team, using Add AI employee. Reconfiguring an existing employee uses the exact same form, and from then on your changes sync to their live setup automatically.

Identity

Set the agent's name, role, and email address (AI employees can have an inbox and send and receive mail). Choose personality traits, for example direct, formal, casual, opinionated, or cautious, to match the role. Write a clear role definition: it is marked optional, but treat it as essential, because it is how the employee understands their mission. Pick an avatar, or a fallback color if you would rather not use one.

Communication and autonomy

• Messaging style: concise, balanced, or detailed.
• Response behavior: acknowledge then work silently (recommended), work first and respond with results, or stream progress updates as they go.
• Autonomy: how much they decide on their own, from figuring things out and only escalating for permissions, money, or strategy, to asking before acting, to always confirming.
• Communication channel: where the employee talks with you, such as Slack, email, WhatsApp, or Discord.
• Output rules: keep reasoning out of messages and avoid posting screenshots. Leaving these on keeps Slack readable when several employees are working at once.

Work schedule and proactivity

Set the timezone and a work mode of 24/7, business hours, or custom. Running 24/7 means more work, which means more model tokens and cost, so business hours can give you more visibility and control. Set the proactivity level (high for strategy leads, low for doers) and the proactive mode (create and do work, or review others' work). Choose a heartbeat interval (15 minutes is recommended) that wakes the employee to check the board, comments, and reviews. Set a blocking timeout, so that if you do not reply within the window, the employee moves the task to Pending and gets on with something else. Finally, set the task creation policy: Requires Approval (the default, where a human approves suggested and finished work) or Trusted (the employee creates, works, and closes tasks on its own, best kept for very well-defined work).

Memory and learning

Memory is what makes an employee feel like more than a chatbot. Choose write-first (recommended and safer, since they record what matters before working) or read-first (faster). Set session retention of 30, 60, or 90 days, which governs active memory; with daily memory files and hybrid memory search on, older context stays available in passive memory rather than being forgotten. Over time this builds real institutional knowledge: the longer an employee runs, the more they understand your business and your preferences.

Scheduled tasks

Add recurring or one-time jobs the employee should run, such as a morning brief, an hourly task check, and an end-of-day report.

Choosing modules

Pick which modules this employee gets, so they receive only the skills their role needs.

Generating and installing

When you generate, HQ produces the configuration files plus setup instructions, with a download tailored to your operating system. A self-managed, technical user downloads the files and follows the setup notes (you can even hand the instructions to your agent to apply them). On a managed plan, cln.work takes it from there. After the first setup, you do not download again: changes you make in the wizard sync to the employee's live instance.

API access

Every AI employee has an API key, and you can create API tokens so other agent platforms can work with HQ through its API.

Related

• Quick start
• Team
• Mission Control