Usage & Cost

ActionFlows makes the cost of automation visible at every stage: while you build, before you run, and after every execution. Instead of discovering a surprise bill at the end of the month, you always know what a flow costs and why.

How usage is measured

ActionFlows tracks:

• Credits: organization balance reserved and settled around a flow run or agent turn
• Tokens: input and output tokens on AI nodes, used for the studio estimate when cost calculation is enabled

Non-AI platform nodes do not add a separate ActionFlows surcharge beyond provider usage on that connection.

Where you see cost

While building

Each AI node shows an estimated cost based on the selected model, your prompt, and the configured max tokens. Change the model or shorten the prompt and the estimate updates instantly.

Before running

When cost calculation is enabled, the studio bottom bar shows a cost estimate for AI nodes. Flow Requirements is a different control: it lists missing credentials, not a dollar total.

After running

Every entry in Run History records the real cost and token usage per node and per run. You can filter, export, and build dashboards on top of this data through the List Runs and Get Run API endpoints.

Estimating monthly cost

Use this formula for any AI node:

Monthly cost =
  runs per month ×
  ( average input tokens × input price per token +
    average output tokens × output price per token )

Example: a support-ticket classifier

10,000 tickets/month
Input:  ~200 tokens per ticket (the email text)
Output: ~20 tokens per ticket (a category label)

Using a fast, low-cost model:
Input:  10,000 × 200 × $0.00000015 = $0.30
Output: 10,000 × 20  × $0.0000006  = $0.12
Total:  ~$0.42 / month

Example: a daily content pipeline

Generate blog draft (frontier model): $0.008
Rewrite for social (fast model):      $0.002
Save to Notion + post to Slack:        free

Per run:  ~$0.010
Monthly (daily): ~$0.30

Controlling cost

• Right-size the model. Use a fast, cheap model (Haiku, GPT-4o mini, Mistral) for classification and routing; reserve frontier models (Claude Sonnet/Opus, GPT-5) for reasoning and content. See the AI Model Selection Guide.
• Run Actionflow Studio before you rely on webhooks or high-frequency schedules so you see real graph cost.
• Cap output length with max tokens: output tokens are usually the most expensive part of a call.
• Batch and cache. Process arrays in one node instead of one run per item, and store reusable results in a database instead of regenerating them.
• Limit loop iterations and add Wait nodes to stay inside provider rate limits.

Next steps

• AI Model Selection Guide: pick the cheapest model that meets your quality bar
• Actionflow Studio: measure real cost on a full graph run before you scale
• Usage: credit charges for one ActionFlow
• Run History: audit actual spend per run
• Credential Management: connect the providers you're billed by