Predictive Prompting

What is predictive prompting?

Predictive prompting automatically suggests relevant follow-up actions after the AI completes a code generation. Instead of thinking "what's next?", you get ready-to-use prompts tailored to your project's current state.

How it helps:

• Save time: No need to think of next steps

• Stay focused: Suggestions keep you moving forward

• Discover features: Learn about capabilities you might not know

• Maintain momentum: Keep building without pausing

• Context-aware: Suggestions match your project's needs

When suggestions appear:

• After AI completes code generation (Build mode)

• After debugging sessions (Debug mode)

• After code reviews (Review mode)

• After any AI response with actionable output

How it works

Predictive prompting uses AI to analyze your project and suggest logical next steps.

The process:

1. You send a prompt in Build, Debug, or Review mode

2. AI generates code and completes the response

3. Predictive AI analyzes:

   • What was just implemented

   • Current project state

   • Session history

   • Common development patterns

4. Suggestions appear as clickable buttons below the AI response

What the AI considers:

• Recent code changes

• Project description and goals

• Session conversation context

• Workspace tier (affects number of suggestions)

• Common next steps for similar tasks

Suggestion format:

• Label: Short description (e.g., "Add error handling")

• Prompt: Full instruction for the AI

• Helper text: Optional context about the suggestion

Suggestion quality

The number of suggestions depends on your workspace subscription tier.

Free tier:

• Gets one carefully selected suggestion

• Focuses on the most valuable next step

• Still highly relevant and actionable

Pro tiers:

• Get up to three diverse suggestions

• Covers different aspects (testing, features, polish)

• More flexibility in choosing your path

Tip: Pro subscribers get more choices, but free tier suggestions are still high-quality and actionable.

Using suggestions

Suggestions appear as purple button chips below AI responses.

Clicking a suggestion

What happens when you click:

1. Prompt auto-fills in chat input

2. You can review the prompt

3. Modify if needed

4. Hit Enter to send

You're in control:

• Review before sending

• Edit the suggestion

• Combine with Quick Actions

• Or ignore and type your own

Example interaction

You send: "Add a login form to the homepage"

AI responds: [Generates login form code]

Suggestions appear:

• 🔘 "Add email validation to the login form"

• 🔘 "Create a signup page"

• 🔘 "Add forgot password link"

You click: "Add email validation to the login form"

Input fills: Prompt ready to send or modify

Modifying suggestions

Common modifications:

• Add specifics: "Add email validation with custom error messages"

• Combine multiple: "Add email validation AND password strength indicator"

• Reference files: "Add email validation to @src/LoginForm.tsx"

Enabling/disabling

Predictive prompting is enabled by default but can be toggled per project.

How to toggle

1. Open your project

2. Go to Project Settings → General

3. Find "Predictive Prompting" section

4. Toggle the switch on/off

5. Saves automatically

When enabled:

• Suggestions appear after AI responses

• Context-aware prompts generated

• Minimal credit usage per response

When disabled:

• No suggestions appear

• Saves credits

• You write all prompts manually

Why you might disable

Disable if you:

• Want to save credits

• Prefer writing your own prompts

• Working on highly specialized features

• Already know exactly what to build next

Keep enabled if you:

• Want productivity boost

• Appreciate guided next steps

• Learning the platform

• Building standard app features

Tip: Leave it enabled until you're very comfortable with App2. The suggestions often reveal helpful features.

Credit usage

Predictive prompting consumes a small amount of credits each time it generates suggestions.

Credit consumption:

• Minimal credit cost per generation

• Much lower than Build mode operations

• Uses App2 credits by default

• Or uses BYOK if enabled (Lifetime tier)

Cost context:

• Generating suggestions costs less than implementing them

• Free tier still gets suggestions despite cost

• Pro tier credit allocation easily covers suggestion costs

• Consider disabling if credits are very limited

BYOK support:

• Lifetime subscribers can use own API keys

• No App2 credit consumption when BYOK enabled

• Same quality suggestions

• Configured in workspace settings

Best practices

When to use suggestions

Use suggestions when:

• You're not sure what to do next

• Building standard app features

• Want to maintain development flow

• Learning common patterns

• Following best practices

Example scenarios:

• Just added login → suggestion: "Add signup page"

• Built data table → suggestion: "Add search functionality"

• Created API endpoint → suggestion: "Add error handling"

When to write your own

Write custom prompts when:

• You have specific requirements

• Building unique features

• Combining multiple changes

• Referencing specific files

• Working from detailed specifications

Combining with other features

Suggestion + Quick Actions:

1. Click a suggestion

2. Add file references with Quick Actions

3. Send combined prompt

Example:

• Suggestion: "Add error handling"

• You add: @src/api/users.ts

• Final prompt: "Add error handling to @src/api/users.ts"

Suggestion + Chat Modes:

• Get suggestion in Build mode

• Switch to Plan mode to outline first

• Then switch back to Build to implement

Examples

After adding a login form

Suggestions might include:

• "Add email validation to the login form"

• "Create a signup page with similar styling"

• "Add 'forgot password' functionality"

After creating a data table

Suggestions might include:

• "Add search/filter functionality to the table"

• "Implement pagination for the data table"

• "Add export to CSV feature"

After building an API endpoint

Suggestions might include:

• "Add error handling and validation"

• "Create frontend component to call this endpoint"

• "Add unit tests for the API endpoint"

After fixing a bug

Suggestions might include:

• "Add defensive checks to prevent this bug"

• "Review similar code for the same issue"

• "Add error logging for debugging"

Troubleshooting

No suggestions appearing

Issue: Predictive prompting enabled but no suggestions show.

Solutions:

1. Check it's enabled in Project Settings → General

2. Verify you're using Build, Debug, or Review mode

3. Wait for AI response to fully complete

4. Refresh page if stuck

5. Check you have available credits

Only getting one suggestion (Pro user)

Issue: Pro tier but only seeing 1 suggestion instead of 3.

Solutions:

1. Verify workspace is actually Pro tier (check settings)

2. AI may determine only one strong suggestion is relevant

3. Some responses naturally have fewer logical next steps

4. This is normal behavior in some cases

Suggestions not relevant

Issue: Suggested prompts don't match what you want to do.

Solution: Write your own custom prompt. Suggestions are helpful hints, not requirements. Ignore and type what you need.

Want to disable for one project only

Issue: Want suggestions on some projects but not others.

Solution: Predictive prompting is per-project. Disable in specific project's settings while leaving it enabled for others.

FAQ

Does predictive prompting cost credits?

Yes, generating suggestions consumes a small amount of credits. The cost is minimal compared to actually implementing the suggested changes.

Can I use suggestions with BYOK?

Yes. If your project uses BYOK (Bring Your Own Keys), predictive prompting also uses your own API keys at no App2 credit cost.

Do I have to use the suggestions?

No. Suggestions are optional productivity aids. Ignore them and type your own prompts anytime.

Can I edit suggestions before sending?

Yes. Clicking a suggestion fills the chat input. Review, modify, or combine with Quick Actions before sending.

Why do suggestions vary in number?

Free tier gets 1 suggestion. Pro tier gets up to 3, but sometimes only 1-2 if the AI determines fewer strong options exist for that specific context.

Can I disable predictive prompting to save credits?

Yes. Toggle it off in Project Settings → General → Predictive Prompting. This prevents suggestion generation and saves credits.

Do suggestions appear in Plan mode?

Not typically. Predictive prompting focuses on actionable code changes (Build, Debug, Review modes). Plan mode is for discussions, so suggestions aren't as applicable.

Are suggestions different for each project?

Yes. Suggestions are context-aware, considering your specific project, recent changes, and current state. Same action on two projects may generate different suggestions.

Can I request different suggestions?

No. Each response generates one set of suggestions. If they're not helpful, write your own prompt or perform another action to get new suggestions.

Related links

• Prompt Enhancement - Improve prompts with AI

• Chat Modes - Build, Plan, Debug, Review, Docs

• Quick Actions - Add context to prompts

• Chat Sessions - Organize conversations

• Plans and Credits - Subscription tiers

• Projects - Project settings

• Getting Started