> ## Documentation Index
> Fetch the complete documentation index at: https://docs.convonexai.tech/llms.txt
> Use this file to discover all available pages before exploring further.

# Core concepts

> Projects, rounds, stories, tiers, fidelity, and what the words in the interface actually mean.

## Project

One app = one project. Each project has its own conversation with the agent, its
own live preview, and its own isolated backend resources. Projects can be renamed,
duplicated, pinned, archived, or deleted from the dashboard, and optionally
assigned to a **client** if you build for others.

## The build pipeline

Every meaningful build follows the same path: the agent **specifies** what you
asked for, **plans** the architecture and breaks it into stories, waits for your
**approval**, then **implements** story by story with verification at each step.

What you see is a simpler vocabulary — the status pill over the preview:

**Planning → Your turn → Building → Testing → Polishing → Ready**

*Your turn* always means the product is waiting on you — usually a plan to review
or a question to answer. It never builds past a decision that belongs to you.

## Story

Plans are broken into **stories** — small, independently buildable and verifiable
pieces ("visitors can submit the contact form", "admins can export bookings").
The Kanban and Tasks tabs track them; each story carries its own promises and its
own verification receipts.

## Round

Work on a project is grouped into numbered **rounds** — one round is one stretch of
building toward a goal. Starting a new round shows you honestly what the previous
round left unfinished, and you can look back through past rounds' history at any
time.

## Promise & proof

At plan time the build makes **promises** — concrete, checkable claims about what
the finished app does. As stories complete, each promise is verified against the
running app and stamped:

| Stamp               | Meaning                                                   |
| ------------------- | --------------------------------------------------------- |
| **PROVED**          | A machine check demonstrated it against the live app.     |
| **CONFIRMED · YOU** | You looked and confirmed it yourself, on a recorded date. |
| **DID NOT WORK**    | It was checked and failed.                                |
| **NOT CHECKED**     | No check has run for this yet.                            |

"Did not work" and "not checked" are deliberately different things — the product
never blends them. The [Proof tab](/verification/proof-tab) breaks the picture
down further, including states where a judgment is yours to make and where the
platform still owes a check. See also
[What "verified" means](/verification/what-verified-means).

## Fidelity: Production vs Prototype

Chosen when you create the project:

* **Production** — the strict path. Verification gates on, a scorecard you can
  hand to a client, no silent mocks.
* **Prototype** — the fast path. Gates off, pieces may be mocked, good for
  exploring an idea. Not meant for handoff.

## Tier: Builder vs Developer

How much machinery you want to see. **Builder** keeps the interface simple.
**Developer** turns everything on: a read-only code view, the agent's memory
lens, the skill shelf, and full confidence signals. Switch any time from the
workspace — your project is unaffected.

## Preview vs Published

Your **preview** is the live development version of the app, visible in the
workspace and at its own preview URL. It sleeps after \~30 minutes idle and wakes
in seconds. **Publishing** is a separate, deliberate step that promotes the app
to a permanent production URL — see [Publish](/shipping/publish).

## Capabilities

The real infrastructure a project can use: authentication, a Postgres database,
file storage, secrets, email, inbound and outbound webhooks, deployment, and its
public URL. Each is provisioned per-project and isolated from every other
project. The [Capabilities](/capabilities) section documents each one.

## Plans & usage

Your account has a plan tier with daily and monthly usage allowances. The usage
page shows live meters; hitting a cap pauses agent work until it resets. See
[Plans & usage](/account/plans-and-usage).