Latenode: how it works, per-second pricing, and who it suits

Latenode in three lines

Latenode is a hosted automation platform: you build scenarios out of nodes, and you can drop in JavaScript or a programmatically driven browser. The vendor, Nocode LTD, founded in 2022, lists its headquarters as Cyprus. It no longer sells credit-based subscriptions: it bills the compute time of each execution, in “CPU seconds”, after a free monthly allowance.

This page analyzes Latenode’s public pricing and documentation as recorded on the date shown. Nobody here has run the product.

What Latenode actually does

A scenario starts from a trigger (a webhook, a schedule, an event in an app) and chains nodes. According to the pricing page, three things set it apart from purely visual tools:

• a JavaScript node in both plans: JavaScript node ("Full-Code Node") included in both plans, including Free;
• a headless browser, meaning a web browser driven by the scenario with no window, used to pull data from a page, fill in a form or take a screenshot; the vendor describes it as “Chromium and Puppeteer-style control”;
• AI building blocks: an “AI Agent” node, RAG storage (documents a model searches for its answers) and a database, all included from the free plan up.

On the catalog, two figures sit side by side: "5,500+ Integrations"; the "About" page also says "1,200+ apps and LLMs in one plan" source.

On triggers, webhooks fire instantly (see the webhook guide). For everything else: 10 minutes (scheduled or polling triggers); 5 or 1 minute as a paid add-on; instant webhooks.

What Latenode bills: compute seconds

Credits are gone. If you read elsewhere that Latenode sells “Micro”, “Starter”, “Grow” or “Prime” plans in credits, that is the old grid: the billing documentation says those plans are archived and visible only on legacy accounts. Holders of an old one-time-payment deal keep their allocation, converted at 1 credit for 30 “CPU seconds”.

The current rule: Latenode meters the compute time of an execution, from start to finish, not its steps. Take a 5-step automation that runs 1,000 times a month. The number of steps changes nothing; run time decides:

Average time per execution	CPU seconds per month	Against the 10,000 free allowance
about 1 second	1,000	a tenth of the allowance
about 10 seconds	10,000	the whole allowance, nothing to pay
about 30 seconds	30,000	20,000 billed seconds, in the first bracket

The same automation would use 4,000 tasks on Zapier (four actions per run) and 1,000 executions on n8n. Here, it all depends on what the steps do. The vendor says so itself: two scenarios with the same number of executions can cost different amounts, because waiting on slow APIs, large files or the headless browser lengthens billed time.

Two details from the documentation: running a single node on the canvas uses nothing, and a nested scenario is only counted within the main scenario’s time. In the run history, the “Duration” column (elapsed time) and the “CPU seconds” column (billed time) are not equal (Latenode documentation, FAQ on the billing unit (accessed November 11, 2026)).

To place this unit next to other tools’ tasks, credits and executions, see the guide to billing units.

The plans at a glance

Table scrolls horizontally →

Latenode plans

Plan	Monthly price	Monthly price, billed annually	Monthly quota included	Limits
Freesource	$0	not recorded	10,000CPU seconds	5 active scenarios, 1 parallel execution, 3 minutes per execution, 3-day history. 10-minute interval, 10 connected accounts, 16 MB files.
Pay as you gosource	$0	not recorded	10,000CPU seconds	No base subscription ("No base fee"): 10,000 free "CPU seconds" every month, then per-second billing in decreasing brackets. Unlimited active scenarios, 5 parallel executions, 10 minutes per execution, 30-day history, 10-minute interval, unlimited connected accounts, 16 MB files. Execution time (up to 60 minutes), interval (5 minutes or 1 minute) and file size (up to 1,024 MB) can be extended with paid add-ons.

Just two plans, with no base subscription:

• Free: 10,000 CPU seconds per month, 5 active scenarios, one execution at a time, up to 3 minutes per execution, 3-day history. No credit card.
• Pay as you go: the same 10,000 free seconds every month, then per-second billing in decreasing brackets (the scale is in the box above). Unlimited active scenarios, 5 parallel executions, 10 minutes per execution, 30-day history.

No trial of a paid plan is mentioned: you move from Free to “Pay as you go” by adding a payment method. Paid monthly add-ons extend the maximum time per execution (up to 60 minutes), shorten the trigger interval or raise the file size limit; their prices are in the notes at the bottom of this page.

A monthly spending cap can be set in the account: once it is reached, new executions fail before they start. The tier table further down repeats the vendor’s own volume examples; the cost calculator compares those volumes with other tools.

Hosting and data

On where the service runs, the list of subprocessors says: Hosting on Microsoft Azure and Amazon Web Services, both listed as "US" in the list of subprocessors source. The pages we read mention neither a choice of EU region nor self-hosting: both points stay “not recorded”, which doesn’t mean they don’t exist.

The documentation states that the database is encrypted at rest, customer instances are logically isolated and audit logs are kept for at least three months (Latenode documentation, customer data protection and encryption (accessed November 11, 2026)).

AI and agents

"AI Agent" node; models used without an API key are billed in "PnP tokens" (1 token = $1), on top of "CPU seconds" source. Tokens are charged at the model provider’s actual usage, plus a margin, on top of CPU seconds. The coding helper, “AI Code Copilot”, is listed as “Trial access” on Free and “Pay per use” on “Pay as you go”.

What the pages don’t say, or say twice

• Invites. The vendor’s pages contradict each other: for “Pay as you go”, the pricing page says you can invite teammates to workspaces, while the billing documentation counts a single workspace invite.
• What happens when the free allowance runs out on the Free plan, short of moving to “Pay as you go”: not recorded.
• Interface languages: not recorded.

Alternatives worth a look

• The closest head-to-head: Latenode vs n8n, two tools that accept code, one billed by the second, the other by the execution and installable on your own server.
• For a unit that is easier to forecast: Activepieces (one credit per run) or n8n.
• The wider picture: workflow automation tools compared, Zapier alternatives and n8n alternatives.