Getting Started
What Amira is, how the Spec → Build → Deploy → Companion pipeline works, and how to start your first project.
Amira turns a plain-language description of the app you need into a governed, deployed application — backed by your data lake and your policies. You describe the outcome; the platform's agents specify, build, deploy, and maintain it, with a human approval gate at every irreversible step.
The pipeline#
Every application moves through four stages, in order. A human approves each irreversible step — nothing is built from an unapproved spec, and nothing is deployed without sign-off.
The Spec Agent interviews you and produces a builder-ready specification: functional and non-functional requirements, acceptance criteria, a capability graph, and any open gaps or decisions. You iterate in plain language until the spec is complete.
Once a spec is approved, the Build Agent writes the application in a sandbox, wiring in the skills and data sources the spec references. Progress is scored against the spec's acceptance criteria on a live compliance matrix.
With approval, the Deploy Agent promotes the built application to a running URL.
Every deployed application gets a companion agent, so other apps and people can call it as a reusable capability.
Start your first project#
From the home screen, choose New Project.
Type a sentence or two describing the app you need. Or import an existing codebase and let Amira reverse-engineer a starting spec from it.
The agent asks focused questions and fills in the specification as you answer. Keep going until it reports the spec is complete — see Working with Specs for how the interview works.
When the spec is ready, route it for approval. Once it's signed, the Build Agent takes over — see E-signature & Approval.
An app can read from and write to your own database or warehouse. Register the source once on the project, paste the credential you were given, and every app built there can reach it — see Data Connections.
You don't have to start from scratch. Point Amira at a Git repository or a folder on your computer and it reverse-engineers a starting specification — requirements, a capability graph, and open gaps — that you then refine like any other spec.
Where to next#
How the interview works, grounding the agent with your own documents, and iterating to a complete spec.
How a finished spec is routed, who can sign, and what an approval records in the audit ledger.
Definitions of the core platform terms — agents, specs, governance, and data.
Quick answers and fixes for the things people hit most often.