The Real Cost Structure of an AI Agent

Almost every cost discussion about AI agents opens with a model price per million tokens, which is the one number that tells you the least. The bill you actually receive is a stack of four things: API calls, infrastructure, the one time build, and the recurring costs nobody put in the estimate. Here is how the stack usually breaks down and which layer is worth attacking first. Where The Money Actually Goes For a typical business agent, a support bot or an internal automation running on a managed platform, monthly operating cost lands between 200 and 1,000 dollars. API calls are 40 to 60 percent of that. Hosting, a vector database for memory, and monitoring share the rest. The spread on either side is wide: a solo developer on open source models and a small VPS can stay under 50 dollars a month, while an enterprise running multi agent systems on frontier models regularly spends 5,000 to 13,000 a month before anyone counts the build. Infrastructure has its own shape. Serverless is the cheapest entry, and a moderate agent handling 10,000 to 20,000 interactions a month usually runs 50 to 200 dollars in compute with no idle charge. Containers on ECS, Cloud Run or Kubernetes cost 100 to 500 and buy persistent connections and steady latency. Self hosted GPU starts around 200 a month for a T4 class instance and passes 1,000 for A100 or H100 class, which only pays off at volumes high enough to amortize it. Vector storage adds 20 to 500, and pgvector on a Postgres you already run removes that line entirely. Model Choice Is A Routing Decision The price spread between tiers is large enough that treating model selection as one global choice is the expensive mistake. Frontier reasoning models sit at the top of the range, mid tier models cost a fraction of that, and the lightweight tier is cheaper again by roughly an order of magnitude. An agent that sends every step to the top tier is paying reasoning prices for string formatting. The fix is routing per step rather than per agent. Classification, extraction, routing and formatting go to the cheap tier. Planning, tool selection under ambiguity, and anything where a wrong answer compounds go to the strong tier. Teams that do this carefully cut the API bill substantially without a quality change users can detect, because the steps that got downgraded were never the ones carrying the quality. Caching And Batching Beat Model Shopping Prompt caching is the single largest reduction available, and it fits how agents work almost perfectly. Every call resends the same system prompt, the same tool definitions and a growing pile of context. Cached input tokens run 80 to 90 percent below standard rates across the major providers, and a caching strategy that respects the cache boundaries routinely takes 50 to 70 percent off total API cost. Batch APIs are the other easy win. Both Anthropic and OpenAI discount batch requests by 50 percent in exchange for higher latency, which costs nothing for background work: content generation, bulk classification, overnight analysis. The rule of thumb is that anything a human is not waiting on should not be paying interactive prices. What It Adds Up To The build sits on top of all of this. A framework based agent is one to four weeks and roughly 2,000 to 15,000 dollars of development. No code platforms come in under 2,000 and trade flexibility for speed. Fully custom work runs 15,000 to 50,000 for mid complexity and considerably higher for enterprise systems with security review and integration work. The pattern worth remembering is that the build is a one time number everyone estimates and the operating cost is a recurring number almost nobody models. Fix the routing, cache the repeated context, batch what can wait, and the recurring number stops growing faster than the usage does. The full breakdown, API pricing through infrastructure and the hidden recurring costs, is in this cost guide.