How bldrAgent works
A deep dive into the mechanics of bldrAgent — how the AI understands your app, generates code, manages context, and keeps everything in sync as you build.
Under the hood: how bldrAgent builds your app
bldrAgent is not a drag-and-drop builder with an AI label on it. It is a genuine AI system that reads, writes, and updates the actual code and data model of your application in real time, based on your natural language instructions.
This guide explains how that process works — so you can work with the system more effectively and understand what's happening when you send a message.
The project context model
Every bldrAgent project maintains a live context — a structured representation of your entire app that the AI reads before responding to each of your messages.
The context includes:
- Pages and components — every screen, section, and UI element in your app
- Data model — your database tables, fields, and relationships
- App logic — form submissions, automations, user roles, conditional rules
- Styles — your brand settings, colour palette, and typography
- Settings — authentication rules, integrations, publish configuration
Because the AI always reads this context, it never misunderstands what already exists. When you say "add a delete button to the bookings table", the AI knows which table you mean, what columns it has, and what deleting a row should trigger — without you having to explain any of it.
The request lifecycle
When you send a message in the bldrAgent chat, here is what happens:
1. Context assembly
The AI collects the current state of your project — pages, data model, logic, and your recent chat history — into a structured prompt.
2. Intent parsing
The AI determines what type of change is being requested:
- UI change — adding, removing, or modifying a component
- Data change — creating or altering a database table or field
- Logic change — adding or modifying app behaviour (forms, automations, conditions)
- Style change — adjusting visual properties
- Multi-step change — a feature that requires changes across two or more of the above
3. Plan generation
For complex requests, the AI generates an internal plan — a sequence of atomic changes needed to deliver the feature. You won't see this unless you ask: "Before you build this, tell me what changes you're planning."
4. Code generation
The AI generates or modifies the underlying code for your app. This includes:
- Component code — the UI layer
- Schema updates — database migrations if new tables or fields are needed
- Logic code — event handlers, API calls, form processing
- Style updates — CSS classes and theme variables
5. Preview update
The canvas in the editor updates in real time as changes are applied. You can preview any screen instantly without deploying.
6. Checkpoint saved
Every successful change is saved as a checkpoint in your version history. You can review, name, or roll back to any checkpoint from Project Settings → History.
How context awareness works
The AI's ability to make precise changes without breaking your existing app comes from its continuous context awareness. A few important properties:
Session context (short-term memory)
Within a single chat session, the AI remembers every message you've exchanged. If you said "add a booking form" at the start and later say "make the form two columns", the AI knows you're talking about the booking form — not some other form.
Project context (structural memory)
Regardless of how long ago you built something, the AI always reads the current state of your project before responding. This is not session-based — it persists across sessions. Come back to a project you haven't touched in weeks and the AI knows every page, table, and rule.
What the AI does NOT remember
- Conversations from other projects — each project's context is isolated
- Anything you deleted — once a component or table is removed, it's gone from context
- External data — the AI does not have access to your live database records, only your schema
AI models
bldrAgent uses large language models (LLMs) to power the AI chat. Different models have different strengths:
| Model tier | Strengths | Best for |
|---|---|---|
| Standard | Fast, cost-efficient | Quick UI changes, styling, simple logic |
| Advanced | More reasoning depth, better at complex architectures | Complex features, multi-table data models, debugging |
| Premium | Highest capability, slower | Enterprise-grade apps, advanced integrations |
You can switch models from Project Settings → AI Model. Model availability depends on your plan — see Billing and plans.
Credits note: Different models consume different amounts of credits per request. Advanced and Premium models cost more per message. See Credits for the full breakdown.
Database generation
When bldrAgent creates or modifies a data model, it uses a migration-based approach:
- The AI defines the schema change (new table, new field, altered relationship)
- A migration is generated and applied to your project's database
- The UI layer is updated to reflect the new data structure
This means your database schema is always in sync with your app's UI. You won't have a form that references a field that doesn't exist.
The canvas and preview
The canvas is a real-time visual representation of your app's current state. It is not a mockup — it renders the actual code the AI has generated.
- Edit mode — shows the component structure with selection handles
- Preview mode — shows the app as your users will see it, with live interactions enabled
You can switch between modes using the toggle in the top bar. Use preview mode to test flows end-to-end before publishing.
Publishing and deployment
When you click Publish:
- Your app is built into an optimised production bundle
- It's deployed to bldrAgent's global CDN
- It goes live at your
yourapp.bldr.appURL (or your custom domain if configured)
Subsequent publishes are incremental — only changed files are redeployed, making updates near-instant.
What bldrAgent is not responsible for
Understanding the boundaries helps you build better:
- Content accuracy — the AI generates functional code and realistic placeholder content, but real content (your actual product descriptions, images, legal copy) must be provided by you
- Third-party service reliability — if you integrate with a payment gateway, email provider, or external API, the uptime and behaviour of those services is outside bldrAgent's control
- Data entered by your users — bldrAgent stores it in your project database, but you are responsible for the data you collect and how you use it
Related
- What is bldrAgent — overview and setup
- Advanced bldrAgent features — power features and customisation
- Using the AI chat — getting the most out of every message
- Writing prompts — prompt patterns that produce the best results