Skip to main content
PuzldAI supports multiple modes for orchestrating AI models to work together on tasks.

​
Available Modes

Compare

Run the same prompt on multiple models and compare responses side-by-side.

Debate

Multi-round discussion where models argue and refine their positions.

Correction

One model produces, another reviews and optionally fixes.

Consensus

Models propose solutions, vote on the best approach, and synthesize a final answer.

​
When to Use Each Mode

ModeBest For
CompareGetting diverse perspectives, choosing between approaches
DebateExploring trade-offs, uncovering edge cases
CorrectionQuality assurance, code review, fact-checking
ConsensusComplex decisions requiring agreement, architectural choices

​
Quick Examples

​
Compare Mode

/compare claude,gemini "Write a function to parse JSON safely"

​
Debate Mode

/debate claude,gemini "Should we use GraphQL or REST for this API?"

​
Correction Mode

/correct claude gemini "Implement binary search"

​
Consensus Mode

/consensus claude,gemini,codex "Design the database schema for a todo app"

​
Background Execution

All multi-agent modes support background execution:
  • Press Escape to hide the view and return to chat (execution continues)
  • Press Ctrl+E to return and check progress
  • Press Ctrl+C to cancel the operation
A status indicator shows progress when running in the background.