This week's chart shows that agentic workloads now consume nearly five times as many tokens as direct human use on OpenRouter. Agentic traffic reached 7.3 trillion tokens on a seven-day rolling average in August. Human traffic only reached around 1.4 trillion.
OpenRouter classifies API keys as agentic, human or mixed using behavioural signals such as tool calls, interaction length and the time between requests. The chart measures tokens rather than individual calls or agents. It covers OpenRouter traffic rather than the whole AI market.
The growth reflects how agents turn one human instruction into many model interactions. An agent may plan a task, retrieve information, call tools, check results and repeat failed steps. Agent swarms extend this process by delegating work to other agents.
Human demand remains the starting point, but it no longer determines the volume of AI activity. Agentic systems multiply each instruction into a growing stream of machine-generated work.