TOPIC AI
Execution Loop the perceive → reason → act cycle that drives autonomous agents
IN 10 SECONDS
The execution loop (also called the agent loop or cognitive loop) is the core cycle of any AI agent: Perceive (gather observations from tools or the environment), Reason (the LLM decides what to do next based on observations and goals), Act (execute the chosen action — call a tool, return a response, or ask for clarification). This loop repeats until the task is complete or a termination condition is met.
GOTCHA Without a maximum iteration limit, an agent can loop infinitely — burning tokens and cost while getting nowhere. Always set a hard cap on loop iterations (e.g., 10-25 steps) and implement a 'stuck detection' heuristic that breaks the loop if the agent repeats the same action twice.
HOW THE PERCEIVE → REASON → ACT CYCLE RUNS
01 Perceive the agent observes its current state — tool outputs, user messages, environment data, and conversation history.
02 Reason the LLM processes all observations and decides the next action: 'I need to query the database to answer this question.'
03 Act the chosen action is executed — a tool call, API request, or final response generation.
04 Loop check was the task completed? If yes, return the result. If no, feed the action's output back into the perceive step and loop.
POKE IT YOURSELF
python -c "while not done and steps < MAX_STEPS: observation = tool.run(action); action = llm.decide(observation); steps += 1" — pseudocode for a basic execution loop
Drill this topic →
~70 sec read