Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.specode.ai/llms.txt

Use this file to discover all available pages before exploring further.

​
How Specode Builds Your App

Specode uses three specialized AI agents that guide you through planning, design, and implementation. You approve each stage before moving to the next — no black-box builds.
1

Describe Your Idea

Start with a simple, plain-English description of what you want to build. You don’t need a detailed spec — just describe the problem you’re solving and who it’s for.Example:
I want to build a mental health platform for young adults.
Patients should be able to find therapists, complete intake,
schedule video sessions, and track mood between sessions.
The Planning Agent takes it from here — it asks clarifying questions (like a good PM would), then produces a scoped roadmap with priorities, user flows, and a clear first build.
You can also upload reference images during planning — screenshots of apps you like, mockups, or wireframes. The AI uses them to understand what you’re going for.
2

Review the Roadmap

The Planning Agent presents a roadmap scoped to an MVP — typically 3-5 screens for the first build. Review it, suggest changes, and approve when you’re happy.You’re in control: Nothing moves forward until you approve. Ask questions, adjust scope, or change priorities before any code gets written.
3

Review the Design

Once the roadmap is approved, the Design Agent shapes how your app looks and feels. It asks about your brand preferences — colors, mood, apps you admire — and produces a design system.If you uploaded reference images earlier, the Design Agent pulls direction from them: color palettes, typography, layout density, and more.Review the design brief and approve it before building begins.
4

Get Your First Build

The Implementation Agent writes your app based on the approved roadmap and design. A working build with 3-5 screens is typically ready in about 10 minutes.Your app comes with:
  • Working authentication with example login credentials for each role — so you can immediately test as a patient, provider, or admin
  • HIPAA-ready infrastructure under the hood
  • A live preview you can interact with right away
5

Iterate and Refine

From here, refine your app through natural conversation. Describe what you want to change and the AI handles it.Example refinements:
In the patient dashboard, show appointments in a 3-column card layout
instead of a list. Add a "Previous Appointments" section below.

Update the color scheme:
- Primary: #007AFF
- Secondary: #5856D6
Keep the existing neutral grays.

Add our branding:
- Logo: [upload file]
- Clinic name: "Mindful Health"
- Update email templates with our logo and colors
Pro Tip: Save major visual changes (branding, color overhauls) for after the core functionality is working. It’s easier to iterate on features without worrying about maintaining custom styling.

​
Building Advanced Features

Once your foundation is in place, you can layer on more complex functionality through conversation:
1

Start with the Core Feature

Describe what you want at a high level. Let the AI ask clarifying questions.Example:
I want automated follow-up reminders when providers send messages
to patients. If a patient doesn't read a message within 2 hours,
send them a reminder.
2

Refine the Details

Add specifics as the AI builds. You don’t need everything upfront — work iteratively.Example:
Add reminder preferences to patient settings:
- Toggle reminders on/off
- Adjustable delay (30 min to 24 hours, default 2 hours)
- Quiet hours setting
3

Add Integrations

For features that need external services, tell the AI what you want and it will guide you through setup.Example:
I want to use CometChat for the reminder messages.
What API keys do I need to add in Project Settings?
The AI tells you exactly which credentials to get. Add them to Project Settings → Integrations, then confirm and the AI handles the rest.
See our Integration Setup Guide for step-by-step instructions on adding credentials.
4

Handle Edge Cases

Once the basics work, refine the behavior.Example:
Add smart reminder logic:
- Cancel reminder if patient reads message before timer expires
- Don't send reminders for automated system messages
- Maximum one reminder per message
- Group reminders if multiple unread from same provider
Important: Complex features rarely work perfectly on the first try. Build incrementally, test each step, and refine based on what you observe.