Credits

Credits are the unit you pay with for AI runs on Nodaro Cloud. Every node that calls a model has a credit price, and you see that price before you run the node. Credits are held when a run starts, charged when the run delivers its result, and returned automatically when the run fails.

What costs credits

• Nodes that run a model: image, video, voice, music, text and analysis nodes. The price depends on the model and its settings.
• Some processing and service nodes have a fixed price. For example, Extract Frame costs 11 credits, and each post with Publish to Social costs 11 credits.
• Nodes that cost nothing: pickers such as Mood, Generation Settings nodes such as Provider, triggers, Save to Storage, Webhook Output, Still to Video and Slideshow.

A Sub-Workflow or Component node costs nothing by itself. The nodes inside it charge their normal price when they run.

See the price before you run

• On the node. The Run button on a node shows the credit price of the run. When one click runs a node several times, for example with the Number of runs setting of the Generate Text - Any LLM node, the button shows the total.
• For AI text nodes, an estimate. AI text and analysis nodes, such as Generate Text - Any LLM, show ~ and a number. This is an estimate of a typical call, because the run is charged for the tokens it uses. See How LLM calls are charged.
• In the settings. The price follows the settings. Change the model, the resolution or the duration, and the price on the Run button changes with it. A Provider node connected to the Settings input of Generate Image or a video node changes the price in the same way, and so does a Duration node on a video node.
• For the whole workflow. Execute workflow shows the total for every node that will run. After runs, the Cost tab of the editor breaks down what the workflow has cost, by node type and model.
• In apps and components. A MiniApp shows its price on the Launch button. A component shows an estimated price in its settings panel.
• In these docs. Every node page and every model page lists its credit prices.

If your balance is too low, the Insufficient Credits dialog shows the credits you need and the credits you have. Its Buy Credits and Upgrade Plan buttons take you to the next step.

What sets the price of a model

A model's price depends on its settings: the resolution or quality of an image, and the duration, resolution and sound of a video. Some video models charge per second. The number in a model table is the price of the default settings, and higher settings cost more.

Compare models of the same kind. A premium image model still costs much less than a premium video model. To choose a model for a job, read Choosing a model.

How a run is charged

Some nodes hold the most a run can cost and charge only what you get. For example, AI Avatar charges for the actual length of the video it makes and returns the rest of the hold when the run finishes.

How LLM calls are charged

Most nodes that call a language model (an LLM) are charged for the tokens the call actually used. Tokens are the small pieces of text a model reads and writes. A call is charged at the rate of the model that ran it. This covers the AI text nodes, such as Generate Text - Any LLM, Generate Script and QA Check. It also covers the vision nodes, such as Describe Image, Describe to Picker and Image Critic. Other examples are the AI judge of Choose Best, the motion graphics and 3D nodes, the AI analysis of Meta Ads, and the description written when you approve an asset's image.

• What is counted. The tokens the model reads, such as your instructions, your request and any images, and the tokens it writes, such as the answer and its reasoning. Each model has one rate for the tokens it reads and a higher rate for the tokens it writes. A premium model costs more per token than an economy model.
• The figure before a run is an estimate. The Run button, the model pickers and POST /v1/credits/model-costs show ~ and a number. This is what a typical call of that node on that model is charged. A long prompt, many images or a long answer cost more than the estimate.
• A hold comes first. When the call starts, Nodaro holds enough credits for the call's first attempt at the longest answer the model may write. If your balance is below the hold, the run is refused before it starts, with the message "Needs up to N credits".
• Then the real charge. After the call, the run is charged what it used. This is usually much less than the hold. A retry or a very long answer can cost more than the hold. The extra is taken from your balance, and any part your balance cannot cover is written off. It is never billed later.
• A small minimum. Every call that the model's maker served has a small minimum charge. A call's total is rounded up to a whole credit once.
• Lanes. By default, a model runs through Nodaro's default service. Some calls run on the maker's own service instead, and are charged at that service's rate. This happens in Advanced mode, when a Claude model has any Reasoning Effort other than Auto, and when the default service fails and the call is tried again. Gemini 4 Argon always runs on Google's own service.
• Reasoning Effort has no price of its own. A higher effort usually writes more reasoning tokens, and those tokens are charged at the model's rate. Max Tokens sets the size of the hold, not the charge.
• Failures. A failed call that the model's maker still billed is charged for that usage, and the error says how much. This covers a refusal, an answer cut off at Max Tokens, and an answer that could not be used. A failure that the maker did not bill is refunded.
• Stopping a run. When you stop a Generate Text - Any LLM run while its answer streams in, the run is charged for the prompt and the text written before you stopped it.
• Image approvals never wait on credits. Approving the image of a character, creature, location or object writes its description with a vision call, charged for the tokens it used. With too little balance for the call, the image is still approved and the description is skipped. You can write it later.

The price key of a call names the feature and the model, for example llm-chat:gemini-3.6-flash, with :direct on the maker's own service. llmPriceKey in the @nodaro/shared/llm-models package builds the key. See Credits for the endpoint.

Content Recipe, Content Ideas and Workflow Copilot are priced differently, as described on their own pages.

When credits come back

• A failed run is refunded automatically. The exceptions are a failure after the model already delivered its result, for example in a post-processing step, and an AI text call that the model's maker billed. See How LLM calls are charged.
• A blocked generation is always refunded. When a model's safety filter blocks a prompt, you get the credits back. Some models retry once automatically, at no extra cost.
• Cancelled training is refunded. Cancelling a character training returns its credits.
• An empty result is not a failure. A run that completes but finds nothing, such as a web search with no results, is charged.

Where credits come from

Subscription credits are spent first each month. Top-up credits are used only after the subscription credits run out. Larger top-ups get a better price per credit, and subscriptions have a lower price per credit at steady volume. Plans and prices are on the pricing page.

The free tier

The free grant works with every model the free tier can run, and it has no daily cap. Results are watermarked until your first purchase. Some settings need a paid plan, for example 4K video output.

Pay-as-you-go

You do not need a subscription to buy credits. Buying any credit pack, or loading any amount, turns on pay-as-you-go:

• Every model is unlocked, results have no watermark, and there is no daily cap.
• The credits stay valid for 12 months from the purchase.
• You spend pay-as-you-go credits through the REST API, the SDK, the CLI and MCP. Using the editor in the browser needs a subscription.
• Your results are public in the Gallery. Private mode is a subscription feature, from the Standard plan up.

API without a subscription shows what a result costs and how to make the first call.

Auto-recharge tops up your balance for you. On the Billing page, set "When my balance drops below" a number of credits, and the dollar amount to load. The amount is charged to your saved card at the same price as a manual load. After three failed charges, auto-recharge turns off until you turn it on again. Every charge emails a receipt, and receipt links appear under Credit Activity.

Check your balance and spending

• The sidebar shows the credits you have left, split into subscription and top-up credits.
• The Billing page shows your balance in each pool, what you used today and this billing cycle, and your plan. Your purchases, such as plan payments and top-ups, are listed under Credit Activity with links to their receipts.
• The Executions list shows the credits each node of a run used. The Cost tab of the editor totals the runs of one workflow.
• From code, read your balance and history with the Credits API, or ask an assistant connected through MCP.

Credits on a self-hosted install

The Community Edition and the Business edition have no credits. They run models with your own provider keys. To use Nodaro Cloud models from a self-hosted install, connect it to your Nodaro Cloud account. Runs that go through the connection are paid with that account's credits. On app.nodaro.ai, Billing › Connected Instances lists each connected install with its spend this month. There you can set a monthly cap of 100 to 1,000,000 credits per install, or disconnect it. See Connect to Nodaro Cloud.