Credits

Credits are the unit of measurement for workflow execution costs in AgentLoops. Each time you run a workflow, credits are consumed based on the nodes used and their configurations.

Why Credits?

We use a credit-based system because it allows us to charge based on what it actually costs to run your workflow. Different operations have different resource requirements:

• AI model calls vary significantly in cost (GPT-4 costs more than Claude Haiku)
• Third-party API calls have associated costs
• Network operations like web fetching consume resources

By using credits, we can pass on savings to you when you use efficient models or free core nodes, while only charging for the resources you actually consume. This means:

• You're not overcharged: Pay only for what you use, not a flat rate that assumes worst-case usage
• We're not overcharging: Our pricing reflects actual infrastructure costs
• Fair and transparent: See exactly where your credits are being spent in every workflow

This approach ensures sustainable pricing that scales with your needs without unnecessary overhead.

How Credits are Calculated

Base Calculation

Every workflow execution has a minimum cost of 1 credit. On top of that, each node in your workflow consumes credits based on its type:

• Core Nodes: Free (0 credits)
• AI Nodes: Variable (2-30 credits per execution, depending on the AI model) - see AI model pricing
• Third-Party Nodes: Variable - see integration pricing
• Web Fetch Nodes: 5 credits per execution

Loop Mode Multiplier

If a node is configured with loop mode enabled, the credit cost is multiplied by the number of iterations:

Node Credits = Base Credits × Number of Iterations

For example:

• An AI node using Claude 3.7 Sonnet (20 credits) with 5 iterations would cost: 20 × 5 = 100 credits
• A Gmail Send node (2 credits) with 10 iterations would cost: 2 × 10 = 20 credits

Total Workflow Cost

The total credit cost for a workflow execution is:

Total Credits = 1 (base) + Sum of all node credits (including loop multipliers)

Estimation vs Actual Usage

Before Execution (Estimation)

Before running a workflow, AgentLoops estimates the credit cost based on:

• Node configurations (including selected AI models)
• Loop mode settings and max iterations

This gives you a preview of potential costs before committing to execution.

After Execution (Actual)

After execution, the actual credit usage is calculated based on:

• Actual number of iterations executed (which may be less than the max if the loop exits early)
• The AI models actually used during execution
• Any nodes that were skipped or failed

Important: For loop modes, we charge for all iterations attempted, not just successful ones. This ensures fair billing even if some iterations encounter errors.

Example Calculation

Let's walk through an example workflow:

1. Start Node (Core): 0 credits
2. Web Fetch Node: 5 credits
3. Extract Data Node (using Claude 3.5 Haiku): 2 credits
4. Agent Node (using Claude 3.7 Sonnet, 3 iterations): 20 × 3 = 60 credits
5. Gmail Send Node: 2 credits

Total: 1 (base) + 5 + 2 + 60 + 2 = 70 credits

For detailed information about credit costs for each node type, see the Credits Costs page.