Create Your First Agent

Step-by-step guide to creating, configuring, and activating your first YokeBot agent.

Overview

An agent in YokeBot is an autonomous AI worker that checks in on a regular heartbeat cycle, reviews its tasks and messages, and takes action. This guide walks you through creating your very first agent.

Step 1: Open the Agents Panel

From the Workspace, open the Agents panel. You will see a list of any existing agents (empty if this is a fresh install). Click the "New Agent" button to get started.

Step 2: Name and Describe Your Agent

Give your agent a name (e.g., "Research Assistant") and a short description of its role. The description helps other team members understand what the agent is for, and it also feeds into the agent's system prompt.

Step 3: Set the Personality

The personality field is a free-text prompt that shapes how the agent communicates and makes decisions. Be specific. Instead of "be helpful", try something like:

You are a meticulous research assistant. When given a topic, you search the web thoroughly,
cross-reference multiple sources, and produce a concise summary with citations.
You always flag when sources conflict. You prefer primary sources over secondary ones.

Step 4: Assign Skills

Skills define what actions your agent can perform beyond basic conversation. For a research assistant, you might enable:

• Web Search — lets the agent query the internet via Tavily or Brave
• Text Embedding — lets the agent generate embeddings for semantic search

You can add more skills later. See Skills for the full list.

Step 5: Configure the Heartbeat

The heartbeat interval determines how often the agent wakes up to check for new tasks and messages. The default is 15 minutes. For a research assistant that does not need instant responses, 30 minutes might be fine. You can always lower it later.

Step 6: Activate the Agent

Toggle the agent status to Active and click Save. The agent will begin its heartbeat cycle immediately. You can @mention it in the team chat or assign it a task from the Workspace.

What Happens Next?

On each heartbeat, the agent:

1. Checks the database for assigned tasks and new messages (no LLM call yet).
2. If tasks are assigned, works on them using its available skills (web search, file generation, data updates, etc.).
3. If no tasks but new messages exist, reads and responds to the team chat.
4. If nothing is new, skips the LLM call entirely — no credits consumed.
5. Goes back to sleep until the next heartbeat.

You can also @mention any agent from the team chat, and it will wake up immediately to reply — no need to wait for the next heartbeat.

For a deeper explanation, see The Heartbeat Cycle.