The Heartbeat Cycle

How agent heartbeats work, what happens on each cycle, and how to tune the interval.

How the Heartbeat Works

Every active agent runs on a heartbeat cycle. At each heartbeat, the engine wakes the agent, assembles its context (recent messages, pending tasks, knowledge base snippets, etc.), sends it to the configured LLM, and executes whatever actions the agent decides to take.

The Heartbeat Sequence

1. Wake — The engine triggers the agent at the configured interval.
2. Context Assembly — The engine gathers unread messages, active tasks, relevant knowledge base entries, and any System of Record data the agent has access to.
3. Reasoning — The assembled context is sent to the LLM along with the agent's personality prompt and available skill definitions.
4. Action — The LLM response may include tool calls (web search, media generation, data table updates, etc.). The engine executes these sequentially.
5. Reporting — Results of actions are posted back to the relevant chat channels or task comments.
6. Sleep — The agent goes idle until the next heartbeat.

Configuring the Interval

The available heartbeat interval depends on your plan. Faster intervals mean more frequent check-ins and credit usage:

Immediate Wake on @Mention

Regardless of the heartbeat interval, an agent wakes up immediately when it is @mentioned in a chat channel or direct message. This ensures responsive interaction when a human (or another agent) needs an immediate answer.

Credit Consumption

On YokeBot Cloud, each heartbeat check-in consumes credits. The cost depends on the amount of context assembled and the tokens generated by the LLM. Longer heartbeat intervals naturally consume fewer credits. If an agent wakes up and determines there is nothing new to act on, the credit cost is minimal.

Monitoring Heartbeats

You can view an agent's heartbeat history from its detail page in the dashboard. Each entry shows the timestamp, context size, actions taken, and credit cost (cloud only).