diff --git a/scripts/token_optimizer.py b/scripts/token_optimizer.py
new file mode 100644
index 00000000..0f94c848
--- /dev/null
+++ b/scripts/token_optimizer.py
@@ -0,0 +1,87 @@
+#!/usr/bin/env python3
+"""
+token_optimizer.py — Token Efficiency & Optimization for the Timmy Foundation.
+
+Analyzes agent logs to identify:
+1. "Chatty" Agents — agents outputting excessive tokens for simple tasks.
+2. Redundant Logs — identifying patterns of repetitive log output.
+3. Tool Output Bloat — identifying tools that return unnecessarily large payloads.
+
+Outputs an "Efficiency Score" (0-100) per agent.
+"""
+
+import os
+import sys
+import glob
+import re
+from pathlib import Path
+from collections import defaultdict
+from typing import Dict, List
+
+AGENT_LOG_PATHS = [
+    "/root/wizards/*/home/logs/*.log",
+    "/root/wizards/*/logs/*.log",
+    "/root/wizards/*/.hermes/logs/*.log",
+]
+
+class TokenOptimizer:
+    def __init__(self):
+        self.agent_stats = defaultdict(lambda: {"tokens": 0, "turns": 0, "tool_calls": 0})
+
+    def estimate_tokens(self, text: str) -> int:
+        # Rough estimate: 4 chars per token
+        return len(text) // 4
+
+    def find_logs(self) -> List[Path]:
+        files = []
+        for pattern in AGENT_LOG_PATHS:
+            for p in glob.glob(pattern):
+                files.append(Path(p))
+        return files
+
+    def analyze_log(self, path: Path):
+        # Extract agent name from path
+        try:
+            parts = path.parts
+            idx = parts.index("wizards")
+            agent = parts[idx + 1]
+        except (ValueError, IndexError):
+            agent = "unknown"
+
+        try:
+            with open(path, "r", errors="ignore") as f:
+                content = f.read()
+                self.agent_stats[agent]["tokens"] += self.estimate_tokens(content)
+                
+                # Count turns (approximate by looking for role markers)
+                self.agent_stats[agent]["turns"] += content.count("[ASSISTANT]")
+                self.agent_stats[agent]["turns"] += content.count("[USER]")
+                
+                # Count tool calls
+                self.agent_stats[agent]["tool_calls"] += content.count("Calling tool:")
+        except Exception as e:
+            print(f"Error analyzing {path}: {e}")
+
+    def run(self):
+        print("--- Token Efficiency Audit ---")
+        logs = self.find_logs()
+        for log in logs:
+            self.analyze_log(log)
+
+        print(f"{'Agent':<20} | {'Tokens':<10} | {'Turns':<6} | {'T/Turn':<8} | {'Efficiency'}")
+        print("-" * 65)
+        
+        for agent, stats in self.agent_stats.items():
+            tokens = stats["tokens"]
+            turns = max(stats["turns"], 1)
+            t_per_turn = tokens // turns
+            
+            # Efficiency score: lower tokens per turn is generally better
+            # Baseline: 500 tokens per turn = 100 score. 2000+ = 0 score.
+            efficiency = max(0, min(100, 100 - (t_per_turn - 500) // 15))
+            
+            print(f"{agent:<20} | {tokens:<10} | {turns:<6} | {t_per_turn:<8} | {efficiency}%")
+
+if __name__ == "__main__":
+    optimizer = TokenOptimizer()
+    optimizer.run()