What an AI Agent Actually Costs to Run Each Month

An AI agent that triages 500 support emails a month costs between roughly $1 and $28 a month in model tokens, depending entirely on which model you point it at. The runtime, if your vendor charges one, is likely free at that volume. The figure that moves your bill is not the price per token — it is how many turns the agent takes, because it re-reads its whole context on every one.

All prices below are from each vendor's own published pricing pages, checked September 2026. Token prices in particular change often, and one of the figures below is already scheduled to double.

The job being priced

Concrete beats generic, so here is a specific agent. It reads an inbound support email, looks the sender up in the CRM, checks their most recent order status, then writes a suggested reply and applies a category label. Three model turns: read and call a tool, receive the tool result and call a second tool, receive that and produce the answer. Five hundred emails a month.

Reasonable token counts for that shape of job:

• System prompt, routing rules and tool definitions: 1,500 tokens, fixed, sent on every turn
• The email itself: 600 tokens
• CRM lookup result: 400 tokens. Order status result: 300 tokens
• Model output: about 150 tokens for each tool call, 350 for the final drafted reply

These are estimates for illustration, not measurements. The last section shows you how to replace them with your own real numbers, which you should do before committing a budget.

The part people forget

The API has no memory. Each request carries the entire conversation so far, and you are billed for all of it again. OpenAI's conversation state guide puts it plainly: "While each text generation request is independent and stateless, you can still implement multi-turn conversations by providing additional messages as parameters." The same page warns that even when you hand OpenAI a response ID instead of the messages, "all previous input tokens for responses in the chain are billed as input tokens in the API."

Anthropic's context windows documentation describes the same mechanic: "as the conversation advances through turns, each user message and assistant response accumulates within the context window, and previous turns are preserved completely." Each turn's input phase "contains all previous conversation history plus the current user message," and "everything in the request counts toward the context window: the system prompt, every message in messages (including tool results, images, and documents), and your tool definitions."

Run the arithmetic on the triage agent and the effect is stark.

Today's published token prices

Per million tokens, standard rates, September 2026. From OpenAI's API pricing page, Anthropic's pricing documentation, and the Gemini API pricing page.

Two footnotes that matter. Google's page states that Gemini 3.8 Flash is $0.75 input and $3.75 output "through December 31, 2026," rising to $1.50 and $7.50 "starting January 1, 2027." If you build a budget on Flash today, write the doubling into next year's forecast now. And Gemini 3.1 Pro charges $4.00 input and $18.00 output once a prompt passes 200,000 tokens — a threshold a long agent loop can reach without anyone noticing.

The monthly bill

At 7,850 input and 650 output tokens per email, 500 emails a month is 3.925 million input tokens and 325,000 output tokens. Multiply through:

Notice that GPT-5.6 Terra, Gemini 3.1 Pro and Claude Sonnet 5 land within 65 cents of each other on this workload. On mid-tier models at this volume, price is a tie and you should choose on output quality and on which vendor's tooling you can live with. Our comparison of AI agent platforms covers that side of the decision.

Caching and batching

The 1,500-token system-and-tools block is identical on all three turns, which is exactly what prompt caching is for. Anthropic's pricing documentation puts a five-minute cache write at 1.25 times the base input price and a cache read at 0.1 times. Apply that to the fixed prefix inside each email's loop and the Claude Sonnet 5 bill drops from $11.10 to about $8.78 a month.

That is a real 21% saving and it is also smaller than people expect, because a five-minute cache does not survive the gap between one email and the next. Caching helps within a loop, not across a month's worth of them.

Batching helps more, if your job can wait. Anthropic's Batch API is a flat 50% off both input and output, which halves Sonnet 5 to about $5.55. That works for a triage queue processed hourly or overnight. It does not work for anything a customer is waiting on.

The runtime charge, where there is one

OpenAI and Anthropic sell you tokens; you run the code yourself, wherever your code already runs. Google's Vertex AI Agent Engine is the one of the three that meters the agent process itself.

From Google's Vertex AI pricing page, September 2026: agent compute is $0.085 per vCPU-hour after a free tier of the first 50 vCPU-hours per month per account, and agent memory is $0.009 per GiB-hour after the first 100 gibibyte-hours. The page states that "all usage is rounded to the nearest second" and that "for Runtime, idle time spent waiting for the next prompt between turns is not billed."

Hosted tools are the charge that does sneak up. OpenAI and Anthropic both price web search at $10.00 per 1,000 calls — a genuine tie. Anthropic's documentation adds that search results "are counted as input tokens," in that turn and in every later turn of the conversation, so a search does not just cost $0.01, it inflates the context you re-read for the rest of the loop. OpenAI separately prices code-interpreter containers per 20-minute session: "$0.03 per GB / $1.92 per 64GB per 20-minute session."

Compare it against the thing it replaces

Three minutes of human handling per email, across 500 emails, is 25 hours. At the US small-business labour cost of $37.36 per hour worked — private establishments with 1 to 49 workers, March 2026, from BLS Employer Costs for Employee Compensation, Table 6 — that is about $934 a month of labour against $2 to $30 of tokens.

That gap is why agents get built, and it is also why the token bill is almost never the number that decides the project. Review time, error handling and the cost of a wrong answer reaching a customer dominate. We work through that calculation properly in how to calculate automation ROI.

Measure your own, because these numbers are not yours

Every figure above rests on assumed token counts. Real prompts differ enormously — a verbose system prompt, a long tool schema, a customer who pastes an entire email thread, or a model that decides on six turns instead of three will all move your bill by multiples, not percentages.

Three ways to get real numbers, in increasing order of usefulness:

1. Rule of thumb. Google's token documentation states that for Gemini models "a token is equivalent to about 4 characters" and "100 tokens is equal to about 60-80 English words." Good enough for a first sanity check, nothing more.
2. Count before you send. Anthropic offers a token counting endpoint that is, in its words, "free to use but subject to requests per minute rate limits." Google's API has a count_tokens method that "returns the total number of tokens in the input only."
3. Log what you were actually billed. Every response from these APIs carries a usage object with the input and output token counts for that call. Log it. Run 20 real emails through the agent, sum the usage across every turn of each one, and take both the average and the worst case. Multiply the average by your monthly volume for the budget, and the worst case by your volume for the number you tell your finance person.

The third method is the only one that catches the failure that actually blows budgets: an agent that loops more times than you designed it to. Token prices are published and stable enough to plan around. Turn counts are not, and they are where the money goes.