FAQ
What exactly does Delta AI produce?
Section titled “What exactly does Delta AI produce?”A system design, a specification set per service, and an implementation per service delivered as a pull request in your own GitHub organisation. Not a code snippet, not a scaffold — a reviewable backend system.
Do I need to know what architecture I want?
Section titled “Do I need to know what architecture I want?”No. You describe the product; the agent designs the architecture and explains its reasoning. If you do have opinions, plan mode lets you review and steer the design before anything is built.
Which languages are supported?
Section titled “Which languages are supported?”TypeScript, Python, Go and Rust. The choice applies to every service in a project.
Where does the code go?
Section titled “Where does the code go?”Into your own GitHub organisation, as pull requests. Delta AI does not host your code and does not require access after the run completes.
Do you deploy it for me?
Section titled “Do you deploy it for me?”No. Delta AI writes the backend; you run it. The output is standard container
images with /healthz and /readyz endpoints and graceful shutdown, so they run
on whatever platform you already use.
What about databases and Kafka?
Section titled “What about databases and Kafka?”You provision them. The generated services expect PostgreSQL and Kafka connection details in their environment; how you run those is your choice.
Can I change the generated code?
Section titled “Can I change the generated code?”Yes. It is your code, in your repository, under your licence. Nothing depends on Delta AI after the pull requests merge.
What happens if a run fails?
Section titled “What happens if a run fails?”The run’s credits are returned. If the failure is retryable, retry it from the change set panel and it continues from where it stopped rather than starting over.
Can I review before anything is built?
Section titled “Can I review before anything is built?”Yes — that is what plan mode is for. Enable it per workflow step and the run pauses to present its plan for approval.
How does merging work?
Section titled “How does merging work?”Once every pull request in a change set is approved, the change set merges automatically as one group. A background process picks up ready change sets every 15 minutes.
What if I only approve some pull requests?
Section titled “What if I only approve some pull requests?”Nothing merges. A change set merges only when every pull request in it is approved.
Can I work on two things at once?
Section titled “Can I work on two things at once?”Yes. A project can have several active sessions, each seeded from the same merged state. They are independent until one lands — if two sessions change the same service, the second merge will conflict and you resolve it on GitHub.
Can I undo a merge?
Section titled “Can I undo a merge?”Not from Delta AI. Once merged, the code is in your default branch and normal git practices apply.
What does it cost?
Section titled “What does it cost?”The first architecture and one service end to end are free — architecture, specification and implementation, including both diagrams. Anything beyond that needs credits. See Pricing.
Why is my run queued?
Section titled “Why is my run queued?”Your available balance is negative, usually because holds from runs already in flight exceed what is left. It is not a plan limit. Add credits or configure automatic top-up, and the run starts on its own. See Credits and billing.
What if the generated code is not what I wanted?
Section titled “What if the generated code is not what I wanted?”Leave the pull request unapproved and say what is wrong in the session. That starts a new conversation in the same session, and the agent revises the same change set. You can iterate as many times as you need.
Comments on the GitHub pull request are for your team — the agent does not read them.
Do credits expire?
Section titled “Do credits expire?”No.
Is my code used to train models?
Section titled “Is my code used to train models?”No. Your code stays in your repository.
Can I use the output commercially?
Section titled “Can I use the output commercially?”Yes. The generated code is yours, with no licence restrictions from Delta AI.