Skip to main content
Use our chatbot assistant named Docs AI to quickly find answers to your questions instead of searching in the Marqeta documentation directly. The embedded language model can answer you in a human-like question/answer format and assist you with building your application. Get the conversation started by asking a question in the text box: feel free to ask as many questions as you like.
Note
We welcome your honest feedback thumbs up/down related to the quality of the results you receive from Docs AI. Input on your experience will be instrumental in making the chatbot even better.

Best practices

Follow the best practices described in this section to level up your chatbot experience with Docs AI.

Aim for clarity

Try to phrase your questions as clearly as possible; adding extra context enables the chatbot to provide more specific answers.
For example, instead of this vague reference to personally identifiable information (PII) requirements in the sandbox:
✗ To use the kyc endpoint in the sandbox, do I have to use a real person?
Try adding specific nouns and verbs to your question:
✓ Can I use test data to send requests to the /kyc endpoint in my sandbox environment?
Instead of this open-ended question:
✗ how I can configure program?
Try adding more descriptive words like “steps” to your question if you are looking for a procedure to follow:
✓ What are the main steps to configuring a card program?

Don’t be afraid to try again

The chatbot works best when you phrase your question using specific verbs and avoid the passive voice. In addition, using punctuation enables the chatbot to better understand ambiguous parts of speech and formulate a more meaningful answer for you.
If the chatbot claims to not know the answer to your question, reword your request and try again. For example, the chatbot did not understand this question:
✗ Hi, Is there a way to set all online refund authorisations to be approved automatically in the testing environment.
By making the following improvements:

- Eliminating the greeting to the chatbot.
- Using US spelling conventions (for example, “authorizations” instead of “authorisations”).
- Rephrasing the question using clearer language and the active voice (such as, “how can I approve”).
- Placing a question mark (”?”) at the end of the query.

The chatbot can better understand the question and provide a detailed, accurate result:
✓ How can I automatically approve all online refund authorizations in the sandbox environment?

Use direct, active language and proper punctuation

We all appreciate a good joke, but avoid asking questions that contain metaphors, irony, or sarcasm. As shown in the example above, questions expressed in the passive voice (“to be approved”) are more difficult for the chatbot to understand than the same questions expressed in the active voice (“to approve”).
For example, entering a metaphorical query like the following will only confuse the chatbot:
✗ How can I get my webhooks to pick up the pace? Right now, they are being sent as slow as molasses!
To yield better results, try rephrasing this question directly and in the active voice:
✓ How can I improve webhook performance?

Provide sufficient context

In each question you ask, give the chatbot sufficient context to answer without referring to any previous questions, because the chatbot does not currently retain the context from one question to the next.
For example, if you ask the chatbot:
✗ What is JIT funding?
and then ask a follow-up question immediately afterwards such as:
✗ How do I implement this?
the chatbot will not realize that the demonstrative pronoun “this” in the second question refers to “JIT funding” from the first question. Instead, you should ask the chatbot:
✓ How do I implement Gateway JIT Funding?

Disclaimer

To help improve our service, this chat may be monitored or recorded by Marqeta and OpenAI. Furthermore, use of this chatbot is subject to Marqeta’s API Terms of Use and OpenAI’s Terms and Policies. The responses generated by this chatbot have not been independently verified by Marqeta. You are encouraged to always validate any information you receive from this or any other source. As part of Marqeta’s commitment to developing AI safely and responsibly, we note that this chatbot is a recent technological advancement. It can occasionally output plausible-sounding, but incorrect or nonsensical answers.