Does my customer need an Octling account?

No. They open the delivery link, enter the code and run one command. No sign-in is needed.

Can a code be used more than once?

Each code completes the task once. Until it has succeeded, the customer can close the page and come back with the same code, or retry after a failure.

Who pays for the AI usage?

For delivery by link and code, you do: runs draw on your credits. For tasks published in the task market, the user pays the price you set plus model usage, and the platform takes a fee from the price.

Can I validate codes with my own server?

Yes, with remote check. Each code is sent to your https URL and a 200 response passes. This requires the Premium plan.

How do I know a delivery really happened?

Every run keeps its status, timings, the full conversation and a completion screenshot, filed in your library under the code.

What does the customer need?

An internet-connected Windows, macOS or Linux computer and the ability to paste one command into a terminal window.

Use cases · Self-service tasks

Send a link and a code. The AI does the setup on your customer’s computer

You sell something that needs setting up on the buyer’s machine: a printer driver, an accounting package, a licensed plugin, a dev environment for a course. Every sale turns into a support session where you remote in and click through the same steps. A self-service task packages those steps once. The customer opens a link, types the code you sent with the order, runs one command, and an AI agent does the job on their computer while they watch.

What your customer sees

They open the delivery link. No account, no sign-in.
They enter the code you gave them.
The page shows one command to run on their Windows, macOS or Linux computer. When it connects, the page turns into a chat.
The agent checks the requirements you listed, does the work and answers questions along the way.
At the end a screenshot is saved, the device is disconnected and the page says the task is complete. They can close the command window.

If they close the page halfway, entering the same code picks up where they left off. If the run fails, “Retry” starts again with the same code.

Setting one up

Create the task. Give it a title and pick the delivery agent; its persona and skills carry over to the delivery page.
Write the instructions. They become the agent’s system prompt, for example: confirm the printer model first, only download drivers from the vendor’s site, print a test page when done.
List the environment. Required conditions end the run as “Conditions not met” if any one fails, before anything is changed. Other conditions are best effort.
Decide how customers talk to it. Up to three suggested prompts they can tap, whether they may type freely, and an optional cap on messages per run. A task can also allow more than one computer per run.
Import codes. Paste them or upload a txt or csv file. Each code completes the task once.
Send the link and one code with each order.

Checking codes against your own system

Instead of an imported list, a task can use remote check: every code a customer enters is sent to an https URL of yours, and a 200 response passes. That lets your shop or licence server decide which order numbers are valid. Remote check requires the Premium plan.

What you see afterwards

Each run is listed with its status (waiting for device, running, done, failed, conditions not met), the device, timings and a completion note. You can open the full conversation read-only, and the completion screenshot is filed in your library under the code, which is useful when a customer says it never worked.

The task market and share links

A task can also be published to the task market after a review. There you set a price in credits per use; the user pays that plus the model usage, the platform takes a fee, and the price is refunded to the user if the run fails or the conditions are not met. For one-off cases, a share link lets someone use a market task without signing in or entering a code, with the cost charged to whoever created the link.

What to know before you rely on it

You pay for the model usage of code-based runs. They draw on your credits. If you run out, customers see “service unavailable” until you top up, so price the service accordingly and consider a message cap.
The customer has to run one command. The page walks them through it, but someone who cannot open a terminal will still need your help.
It is an AI following your instructions, not a fixed installer. Unusual machines can fail. Write required conditions for what you know breaks the job, and test the task on a few different computers before selling it.
You can read the conversation. Tell your customers that, the same as you would for a recorded support session.
The agent works with the rights of the user who ran the command. Steps that need administrator rights only work if that user has them.

Questions

Does my customer need an Octling account?: No. They open the delivery link, enter the code and run one command. No sign-in is needed.
Can a code be used more than once?: Each code completes the task once. Until it has succeeded, the customer can close the page and come back with the same code, or retry after a failure.
Who pays for the AI usage?: For delivery by link and code, you do: runs draw on your credits. For tasks published in the task market, the user pays the price you set plus model usage, and the platform takes a fee from the price.
Can I validate codes with my own server?: Yes, with remote check. Each code is sent to your https URL and a 200 response passes. This requires the Premium plan.
How do I know a delivery really happened?: Every run keeps its status, timings, the full conversation and a completion screenshot, filed in your library under the code.
What does the customer need?: An internet-connected Windows, macOS or Linux computer and the ability to paste one command into a terminal window.