Using Polydev with OpenAI Codex CLI

$ codex mcp list

Available MCP Servers:
──────────────────────────────────────────────────
  polydev     ✓ enabled   Multi-model AI perspectives
  exa         ✓ enabled   Web search
  github      ✓ enabled   GitHub integration
  supabase    ✓ enabled   Database queries
  vercel      ✓ enabled   Deployment management
  memory      ✓ enabled   Knowledge graph
  filesystem  ✓ enabled   File operations

polydev.get_perspectives success in 39.15s:

🟢 **Local CLI Response** (gemini-3-flash-preview)
I will search the codebase for useEffect usage to provide relevant examples.
*Latency: 17582ms | Tokens: 66*

---

🟢 **Local CLI Response** (claude-haiku-4-5)

## React useEffect Infinite Loops

useEffect infinite loops happen when the dependency array causes the
effect to repeatedly trigger itself. Here are the main causes:

### 1. Objects and Functions in Dependencies
React compares dependencies using Object.is(), which checks reference
equality, not value equality.

// ❌ Infinite loop
useEffect(() => {
  setUser({ name: 'John', age: 30 });
}, [user]); // Creates a new object each time

// ✅ Correct
const initialUser = { name: 'John', age: 30 };
useEffect(() => {
  setUser(initialUser);
}, []); // Only runs once

### Key Takeaway
The dependency array tells React "only run this effect when these
values change." If your effect updates one of those dependencies,
it triggers itself again → infinite loop.

*Latency: 31520ms | Tokens: 1150*

# Quick multi-model consultation
alias poly='codex exec "Use polydev:"'

# Usage:
# poly "best way to implement caching?"
# poly "review this auth code: $(cat auth.ts)"

# Auto-Invoke Polydev for Multi-Model Perspectives

When the user encounters any of these situations, automatically
use the polydev MCP tool (get_perspectives) to consult multiple
AI models before responding:

1. User says "stuck", "confused", "not sure", or "can't figure out"
2. Comparison questions: "Should I use X or Y?"
3. Security or code review requests
4. After 2-3 failed debugging attempts
5. Architecture or design decisions
6. Performance optimization questions

Example: If user says "I'm stuck on this auth bug", call
polydev.get_perspectives with the problem context before answering.