Tell Air Pipe what you want in plain language and it writes the config — validated against the live schema, grounded in real marketplace patterns, ready to deploy. Then edit it by hand or visually whenever you like.
From a prompt to a deployable config
Air Pipe's AI turns a sentence into a working configuration: interfaces, actions, database queries, auth and assertions. It retrieves over the marketplace and the config schema, so the output uses real patterns — not hallucinated fields.
Schema validation with auto-repair before you ever deploy
Grounded in your database schema and published packs
Edit the result as readable YAML or in the visual editor
No black boxes
Unlike a chat that returns opaque automation, Air Pipe returns config you own and can read. Every generated endpoint still ships tracing, metrics and OpenAPI docs automatically — so you can see exactly what it does the moment it runs.
Three ways to build, one runtime
Prompt it, write the config by hand, or fork a marketplace pack. All three produce the same declarative config the engine runs — so you're never locked into one authoring style.
Frequently asked questions
Do I have to keep using AI to maintain the config? No. The AI writes a config, but the config is the product — readable YAML you own. Edit it by hand, in the visual editor, or ask for another pass. Nothing is locked behind the prompt.
How does Air Pipe avoid hallucinating config fields? Generation is grounded in the live config schema and published marketplace packs, and the result is validated with auto-repair before you deploy. Invalid fields are caught rather than shipped.
Can it see my database schema? Yes — generation can be grounded in your database schema, so the queries it writes reference tables and columns that actually exist.