A: Hi barefootmedicine
Let’s break down how Release0 and OpenAI costs work so you can plan your own pricing.

1 - Two Separate Meters: Release0 vs. OpenAI
• Release0 submissions
A “submission” is an interaction with your agent where at least one interaction has been registered from your user, for example, a chat message, a form field entry, or a rating.
• If a user interacts with 50 blocks in your agent flow, that counts as 1 submission.
• If they open the agente and do now interact at all there is no submission so it does not count.
• OpenAI token usage
Every time an AI block runs, OpenAI charges for tokens (roughly chunks of text) used to send the prompt and return the answer.
• The cost depends on the model (e.g., GPT-4.1 vs. GPT-4-mini) and on how long the conversation and the prompt history are.
• This cost is completely separate from Release0 submissions.
• Release0 doesn’t add any markup: you pay OpenAI directly under your own API key (BYOK).

2- Why Testing Burns Credits Fast

When you create and test agents, every test conversation is a real API call to OpenAI and counts toward your token usage.
• If you retry many times or use long prompts, tokens add up quickly.
• Using a less expensive model (like gpt-4.1-mini or gpt-4.1-nano) for prototyping will reduce costs.
• On our end (Release0) does not count any use since the simulator (side panel) does not incur in usage, only using the published URL will count towards your use quota.

3 - Planning for Paying Customers

When customers use your agent, their questions keep generating token charges in your OpenAI account.
• Your Release0 submissions per month (for example 12 000 on Tier 5) limit how many user inputs you can handle per billing cycle. You you will be able to attend to 12.000 people in one month on your Tier5 (if a person interacts on 2 occasions that is counted as 2 interacions)
• Your OpenAI bill depends on how many questions they ask and how large the answers/prompts are.

To cover your costs you can:
• Estimate an average number of questions per user per month and the average token size of an answer.
• Multiply by the model’s per-token cost to calculate your expected OpenAI expense.
• Price your own subscription plans so that the income comfortably covers those expenses.

Example rough math (for illustration):
• Suppose a user averages 50 questions/month.
• Each question/answer pair uses ~1 000 tokens total.
• On GPT-4.1-mini ($0.0006/1 000 tokens), that’s about $0.03 per user/month in OpenAI cost.

4 - Controlling or Limiting Usage

Release0 lets you track and limit usage in your flow:
• Store a counter in a variable for each user.
• Add notification so you can suspend the agent if your OpenAI budget surpasses a certain amount (you could also limit on OpenAI side).

Key Takeaways
• Release0 submissions ≠ OpenAI tokens. Submissions are about user inputs for Release0 analytics and quotas.
• OpenAI costs scale with actual tokens used, and you pay those directly to OpenAi.
• To price your product, estimate average usage and token size, then add a safety margin.
• You can build limits into your agent to control free or paid tiers.

Docs for reference:
• Submissions and billing: https://docs.release0.com/submission/overview#how-to-handle-submissions-in-release0
• OpenAI pricing: https://openai.com/pricing

With these numbers you can confidently design subscription plans that cover costs and stay profitable.