How Onyx works
One platform for the work between an idea and a release.
Onyx connects application development, Agent Research, databases, and AI. Start with the part you need and bring the rest into the same workflow as your application grows.
- 01
Start with a working application
Choose an organization, then start a web or mobile app, select a sample, or import a GitHub repository. New apps include a database by default. Samples get their own database copy; advanced options let other projects use an existing database or add one later.
Generated apps start with local Git. Connect GitHub when you need private imports, pulls, pushes, or draft pull requests.
Create your Workspace - 02
Investigate and prepare with Agent Research
Ask a question about your data or describe a task. The agent inspects the schema, reads relevant records, runs analysis, and returns findings with supporting evidence. Grant actions when you want it to create resources or make changes.
Create dashboards, documents, scripts, and more with specific permissions. Choose the available model and thinking level for the task.
Put Agent Research to work - 03
Build through conversation, review the code
Give the Workspace agent a feature or fix to work on. Follow its progress, inspect the changed files and diffs, and continue with a follow-up. Commit your reviewed work to preserve a version you can release.
The checkout and task history persist when you leave the page. Branches, checkpoints, commits, and connected GitHub workflows keep changes reviewable.
Build with the agent - 04
Try the app and connect its services
Starter setup prepares the database, managed credentials, initial schema, and private browser preview. Test different viewport sizes, inspect startup logs, and rebuild when dependencies or compiled code change.
Private starter previews use your Workspace session. Configure Firebase for deployed apps, and review later schema changes in Environments before applying them.
Prepare your preview - 05
Add the intelligence your application needs
Use Onyx agents for reusable workflows and Model Builder to train neural networks from saved queries. Call published predictions and data APIs from your application through the client SDKs.
Your data model, saved queries, and published model versions give application features a shared foundation.
Explore AI and predictive models - 06
Publish, promote, and keep improving
Publish a committed application version, inspect build and readiness results, and manage the schema changes for each environment. Promote the tested image to production once its database and configuration are ready.
Releases record the code, image, configuration, and migration versions. Roll back to a compatible application image when needed; database migration recovery is separate.
Publish an application release - 07
Build for Android and iOS
For an Expo mobile application, connect a deployed backend and your Expo project, configure native signing, and build a committed version. Install device previews through their links or QR codes, then create store builds for Google Play internal testing or Apple TestFlight.
Mobile builds require runtime support and the relevant developer accounts. Complete store listings, testing requirements, review, and public release in the store consoles.
Build and test a mobile app
For example
Build a support portal around your customer data.
Investigate a growing backlog with Agent Research. Build a ticket queue and customer views in Workspaces. Connect sign-in, review access rules, and try the app in preview. Publish a staging version, then promote it when production is ready.