Skip to main content

Execution Time Pricing

Alongside token charges, AI nodes and paid built-in tools accrue a small charge for the time they spent running. It's the component people don't expect, and the one that explains why two flows with identical model usage can cost different amounts.

Every other node pays a fixed fee instead of per-second time, plus a $0.001 step for each full 30 seconds it runs — see Fixed-fee nodes below.

Why running time costs money​

Token prices pay the AI provider. Execution time pays for the infrastructure your flow actually runs on — the machines holding your flow in memory, waiting on APIs, and doing everything between the model calls.

An Agent that waits nine seconds for a slow model, or for a tool calling a slow third-party API, occupied a worker for nine seconds however few tokens it used. That's what this charge covers.

How execution time is measured​

Per metered node, in wall-clock seconds, from when the node starts to when it finishes. Summed across every AI node and paid tool in the run.

Wall clock is the important part. Time spent waiting counts: a model taking its time, a large document being embedded, a tool an Agent is waiting on. The flow isn't doing anything during the wait, but the capacity is still reserved.

An Agent's time already includes the tools it calls, so those tools never pay time of their own — not even a metered one like web search.

Per-minute rate​

The rate appears in the Exec (per min) column of the Pricing tab — at the time of writing, $0.12 per minute, or $0.002 per second.

Some arithmetic to calibrate:

FlowMetered execution timeExecution charge
Start → Agent → Direct Reply~3 seconds in the Agent~$0.006, plus the Direct Reply's $0.001 fee
Retriever + Agent calling two tools~10 seconds~$0.02, plus each tool's fee
Multi-agent flow with long replies~30 seconds~$0.06
HTTP node waiting 2½ minutes on an APInone — fixed fee$0.006: the $0.001 fee and five 30-second steps

Small per message. Meaningful at ten thousand messages, and meaningful in a batch run that fires a thousand of them at once.

Fixed-fee nodes: 30-second steps​

Nodes and tools that aren't AI models or paid tools — Condition, Filter, HTTP, Custom Function, Direct Reply, Google Sheets, messaging tools and the rest — have no per-second charge. Each run pays its fixed fee, and one $0.001 step for each full 30 seconds it runs:

Node runs forCharge on a $0.001 node
Under 30 seconds$0.001
30–59 seconds$0.002
60–89 seconds$0.003

A Wait node, a tool called by an Agent, and a tool that runs another flow or agent (Agent as Tool, Chatflow Tool, Chain Tool) never pay steps. In the cost breakdown, the steps are part of Execution Time and the fees are Node Fees.

What makes a flow run longer​

In roughly descending order of impact:

  • Slow external calls inside an Agent. A tool the Agent calls that waits on a third party adds that wait to the Agent's execution time. Nothing about the flow is inefficient — the other end is slow. A standalone HTTP node or tool pays only a 30-second step for each full 30 seconds it waits.
  • Long model responses. Generating 800 words takes measurably longer than generating 80.
  • Node count. Every AI node adds its own time, and every other node adds its $0.001 fee. A twelve-node flow that a six-node flow could do pays twice for both.
  • Large retrievals. Fetching and embedding more passages takes longer than fetching fewer.
  • Iteration and loops. Iteration and Loop multiply everything inside them.
note

A Wait node costs a flat $0.001, and its duration is free. The wait suspends the run rather than occupying a worker, so it pays no execution time and no 30-second steps — a one-minute wait and a full-day wait cost the same. A single Wait node can't exceed 24:59:59.

Finding the expensive flows​

The Usage Dashboard's distribution view breaks cost down by flow and by node. When a flow costs more than it looks like it should, the node-level breakdown usually shows one step accounting for most of it — usually an AI node with a long prompt, a long reply, or a slow model.