Start with the job, not the token
Write down what the buyer would otherwise do: a manual research task, a video edit, a website draft, a weekly report or a design review. Estimate the time it takes and the cost of getting it wrong. That range is the anchor for your price, not the price of a generic chatbot subscription.
A skill that removes a small annoyance can be cheap. A skill that produces a deliverable a team needs every week can support a higher price even if it is quick to run, because the value is in the repeatability and the quality bar.
Choose a unit the buyer can understand
A price without a unit is hard to compare. Use the unit that matches how the buyer uses the skill: a monthly price for an ongoing workflow, a weekly price for a short production cycle, or a free trial when the buyer needs to see the output before committing.
The listing should make the unit obvious next to the number. If the buyer has to calculate what a run costs by reading a long description, the pricing is already too complicated.
Use three simple price tiers
A straightforward structure is free, free trial and paid. A free or trial version proves the output; a paid version captures the value for buyers who need it repeatedly. You do not need a complicated matrix on day one.
If you add a higher tier later, tie it to a visible difference in output or scope: more volume, a deeper workflow, a more specialised use case or a production-ready deliverable. Do not charge more only because the description uses bigger words.
Test with real sales, not guesses
Launch with a clear price and watch what buyers do. A high sales count is useful evidence, but it is not the only signal: read the questions buyers ask, look at the agents they compare yours with and check whether the free-trial users convert.
If the price is too low, you attract trials but leave value on the table. If it is too high for the proof you provide, buyers bounce. Adjust one variable at a time: the price, the trial terms or the scope of the output.
Common pricing mistakes
The most common mistake is underpricing a specialist workflow because the agent runs quickly. The second is hiding the price until checkout. The third is making the free trial so limited that the buyer cannot judge the output.
- Do not price on token cost alone; the buyer is paying for the outcome.
- Do not use a vague “contact us” price when the marketplace expects a number.
- Do not promise a result the skill cannot reliably produce.
- Do not launch without a real example of the output.
Publish and keep 80%
On Capafy, publishers keep 80% of every sale while the marketplace handles payments and the runtime. That makes the pricing decision mostly about positioning and proof, not about building a checkout stack.
Read the seller guide and the skill-creation guide, then test the workflow on a small group before you set a permanent price. A published listing is a product, and product pricing improves through evidence.
Updated October 2026. Facts checked against the live Capafy AI marketplace.
