diff --git a/bin/request_log.py b/bin/request_log.py
new file mode 100644
index 00000000..e410d455
--- /dev/null
+++ b/bin/request_log.py
@@ -0,0 +1,255 @@
+#!/usr/bin/env python3
+"""
+Request Log Telemetry — "Verify What Actually Happened"
+
+Issue #446: [P2.5] request_log Telemetry Table
+
+Every agent writes a row to request_log for every inference call.
+No exceptions. No summarizing. Actual rows.
+
+This module provides:
+  - log_inference(): write a telemetry row
+  - query_requests(): read recent telemetry
+  - did_agent_call_provider(): answer verification questions
+
+Database: ~/.local/timmy/request_log.db
+Override via REQUEST_LOG_PATH environment variable.
+"""
+
+import os
+import sqlite3
+import sys
+import json
+from datetime import datetime, timezone, timedelta
+from pathlib import Path
+from typing import Optional, Dict, Any, List
+
+# Default DB location (matches ansible group_vars/wizards.yml)
+DEFAULT_DB_PATH = Path.home() / ".local" / "timmy" / "request_log.db"
+
+
+def get_db_path() -> Path:
+    """Return the configured request_log database path."""
+    env_path = os.environ.get("REQUEST_LOG_PATH")
+    if env_path:
+        return Path(env_path).expanduser()
+    return DEFAULT_DB_PATH
+
+
+def ensure_db() -> Path:
+    """
+    Ensure the database and schema exist.
+    Creates the DB and schema if missing.
+    Returns the DB path.
+    """
+    db_path = get_db_path()
+    db_path.parent.mkdir(parents=True, exist_ok=True)
+    
+    if not db_path.exists():
+        # Create with schema
+        schema = """
+CREATE TABLE IF NOT EXISTS request_log (
+    id INTEGER PRIMARY KEY AUTOINCREMENT,
+    timestamp TEXT NOT NULL DEFAULT (datetime('now')),
+    agent_name TEXT NOT NULL,
+    provider TEXT NOT NULL,
+    model TEXT NOT NULL,
+    endpoint TEXT NOT NULL,
+    tokens_in INTEGER,
+    tokens_out INTEGER,
+    latency_ms INTEGER,
+    status TEXT NOT NULL,
+    error_message TEXT
+);
+
+CREATE INDEX IF NOT EXISTS idx_request_log_agent
+    ON request_log (agent_name, timestamp);
+
+CREATE INDEX IF NOT EXISTS idx_request_log_provider
+    ON request_log (provider, timestamp);
+
+CREATE INDEX IF NOT EXISTS idx_request_log_status
+    ON request_log (status, timestamp);
+"""
+        conn = sqlite3.connect(str(db_path))
+        conn.executescript(schema)
+        conn.commit()
+        conn.close()
+    
+    return db_path
+
+
+def log_inference(
+    *,
+    agent_name: str,
+    provider: str,
+    model: str,
+    endpoint: str,
+    tokens_in: Optional[int] = None,
+    tokens_out: Optional[int] = None,
+    latency_ms: Optional[int] = None,
+    status: str = "success",
+    error_message: Optional[str] = None,
+    db_path: Optional[Path] = None,
+) -> Optional[int]:
+    """
+    Log a single inference request to the request_log table.
+    
+    Args:
+        agent_name: Name of the agent making the call
+        provider: Provider name (anthropic, openrouter, ollama, etc.)
+        model: Model identifier
+        endpoint: API endpoint called
+        tokens_in: Input token count (optional but recommended)
+        tokens_out: Output token count (optional but recommended)
+        latency_ms: Latency in milliseconds (optional but recommended)
+        status: One of 'success', 'error', 'timeout', 'fallback'
+        error_message: Error text if status is error/timeout
+        db_path: Override DB path (for testing)
+    
+    Returns:
+        Row ID if inserted, None on failure
+    """
+    db = Path(db_path) if db_path else get_db_path()
+    
+    try:
+        # Ensure DB exists
+        if not db.exists():
+            ensure_db()
+        
+        conn = sqlite3.connect(str(db))
+        cursor = conn.cursor()
+        cursor.execute("""
+            INSERT INTO request_log
+                (timestamp, agent_name, provider, model, endpoint,
+                 tokens_in, tokens_out, latency_ms, status, error_message)
+            VALUES (datetime('now'), ?, ?, ?, ?, ?, ?, ?, ?, ?)
+        """, (
+            agent_name, provider, model, endpoint,
+            tokens_in, tokens_out, latency_ms, status, error_message
+        ))
+        row_id = cursor.lastrowid
+        conn.commit()
+        conn.close()
+        return row_id
+    except Exception as e:
+        # Never break production — swallow errors for telemetry
+        return None
+
+
+def query_requests(
+    *,
+    agent_name: Optional[str] = None,
+    provider: Optional[str] = None,
+    model: Optional[str] = None,
+    hours: int = 1,
+    status: Optional[str] = None,
+    limit: int = 100,
+    db_path: Optional[Path] = None,
+) -> List[Dict[str, Any]]:
+    """
+    Query recent inference logs.
+    
+    Args:
+        agent_name: Filter by agent name
+        provider: Filter by provider
+        model: Filter by model
+        hours: Lookback window (default 1 hour)
+        status: Filter by status ('success', 'error', etc.)
+        limit: Max rows to return
+        db_path: Override DB path
+    
+    Returns:
+        List of matching records as dicts
+    """
+    db = Path(db_path) if db_path else get_db_path()
+    
+    if not db.exists():
+        return []
+    
+    conditions = ["timestamp > datetime('now', '-' || ? || ' hours')"]
+    params = [hours]
+    
+    if agent_name:
+        conditions.append("agent_name = ?")
+        params.append(agent_name)
+    if provider:
+        conditions.append("provider = ?")
+        params.append(provider)
+    if model:
+        conditions.append("model = ?")
+        params.append(model)
+    if status:
+        conditions.append("status = ?")
+        params.append(status)
+    
+    where_clause = " AND ".join(conditions)
+    
+    try:
+        conn = sqlite3.connect(str(db))
+        conn.row_factory = sqlite3.Row
+        cursor = conn.cursor()
+        cursor.execute(f"""
+            SELECT * FROM request_log
+            WHERE {where_clause}
+            ORDER BY timestamp DESC
+            LIMIT ?
+        """, tuple(params) + (limit,))
+        
+        rows = [dict(row) for row in cursor.fetchall()]
+        conn.close()
+        return rows
+    except Exception:
+        return []
+
+
+def did_agent_call_provider(
+    agent_name: str,
+    provider: str,
+    hours: int = 1,
+    min_success_count: int = 1,
+    db_path: Optional[Path] = None,
+) -> bool:
+    """
+    Answer: "Did agent X actually call provider Y in the last N hours?"
+    
+    Returns True if agent made at least min_success_count successful calls.
+    """
+    rows = query_requests(
+        agent_name=agent_name,
+        provider=provider,
+        hours=hours,
+        status="success",
+        db_path=db_path,
+    )
+    return len(rows) >= min_success_count
+
+
+def get_recent_activity_summary(hours: int = 1) -> Dict[str, Any]:
+    """Get aggregate statistics for recent activity (uses v_recent_activity view if available)."""
+    db = get_db_path()
+    if not db.exists():
+        return {"error": "Database not found"}
+    
+    try:
+        conn = sqlite3.connect(str(db))
+        conn.row_factory = sqlite3.Row
+        cursor = conn.cursor()
+        
+        # Try the view first
+        try:
+            cursor.execute("""
+                SELECT agent_name, provider, model, status, 
+                       COUNT(*) as call_count, AVG(latency_ms) as avg_latency
+                FROM request_log
+                WHERE timestamp > datetime('now', '-' || ? || ' hours')
+                GROUP BY agent_name, provider, model, status
+            """, (hours,))
+            rows = [dict(row) for row in cursor.fetchall()]
+            conn.close()
+            return {"by_agent_provider": rows}
+        except Exception:
+            conn.close()
+            return {"error": "query failed"}
+    except Exception:
+        return {"error": "db error"}
diff --git a/tests/test_request_log.py b/tests/test_request_log.py
new file mode 100644
index 00000000..f09e34a1
--- /dev/null
+++ b/tests/test_request_log.py
@@ -0,0 +1,260 @@
+#!/usr/bin/env python3
+"""
+Tests for bin/request_log.py — Request Log Telemetry.
+Issue #446: [P2.5] request_log Telemetry Table — Verify What Actually Happened
+"""
+
+import json
+import sqlite3
+import tempfile
+import unittest
+from pathlib import Path
+from datetime import datetime, timezone, timedelta
+
+import sys
+import os
+sys.path.insert(0, str(Path(__file__).resolve().parent.parent / "bin"))
+from request_log import (
+    get_db_path,
+    ensure_db,
+    log_inference,
+    query_requests,
+    did_agent_call_provider,
+    get_recent_activity_summary,
+)
+
+
+class TestRequestLog(unittest.TestCase):
+    def setUp(self):
+        """Create a temporary test database for each test."""
+        self.tmpdir = Path(tempfile.mkdtemp())
+        self.db_path = self.tmpdir / "test_request_log.db"
+        # Patch the module's db path by overriding env
+        self.original_env = os.environ.get("REQUEST_LOG_PATH")
+        os.environ["REQUEST_LOG_PATH"] = str(self.db_path)
+        # Clear any cached state
+        if self.db_path.exists():
+            self.db_path.unlink()
+    
+    def tearDown(self):
+        """Clean up test database."""
+        if self.db_path.exists():
+            self.db_path.unlink()
+        # Restore env
+        if self.original_env is not None:
+            os.environ["REQUEST_LOG_PATH"] = self.original_env
+        else:
+            os.environ.pop("REQUEST_LOG_PATH", None)
+    
+    def test_ensure_db_creates_schema(self):
+        """ensure_db() creates the database with correct schema."""
+        db = ensure_db()
+        self.assertTrue(db.exists())
+        
+        # Check table exists
+        conn = sqlite3.connect(str(db))
+        cursor = conn.cursor()
+        cursor.execute("SELECT name FROM sqlite_master WHERE type='table' AND name='request_log'")
+        result = cursor.fetchone()
+        self.assertIsNotNone(result, "request_log table should exist")
+        # Check indexes exist
+        cursor.execute("SELECT name FROM sqlite_master WHERE type='index' AND name='idx_request_log_agent'")
+        self.assertIsNotNone(cursor.fetchone())
+        conn.close()
+    
+    def test_log_inference_inserts_row(self):
+        """log_inference() inserts a complete row."""
+        row_id = log_inference(
+            agent_name="test-agent",
+            provider="anthropic",
+            model="claude-sonnet-4-20250514",
+            endpoint="/v1/messages",
+            tokens_in=123,
+            tokens_out=456,
+            latency_ms=1500,
+            status="success",
+        )
+        self.assertIsNotNone(row_id, "Should return row ID")
+        self.assertGreater(row_id, 0)
+        
+        db = get_db_path()
+        conn = sqlite3.connect(str(db))
+        conn.row_factory = sqlite3.Row
+        cursor = conn.cursor()
+        cursor.execute("SELECT * FROM request_log WHERE id = ?", (row_id,))
+        row = cursor.fetchone()
+        conn.close()
+        
+        self.assertEqual(row["agent_name"], "test-agent")
+        self.assertEqual(row["provider"], "anthropic")
+        self.assertEqual(row["model"], "claude-sonnet-4-20250514")
+        self.assertEqual(row["endpoint"], "/v1/messages")
+        self.assertEqual(row["tokens_in"], 123)
+        self.assertEqual(row["tokens_out"], 456)
+        self.assertEqual(row["latency_ms"], 1500)
+        self.assertEqual(row["status"], "success")
+    
+    def test_log_inference_minimal_fields(self):
+        """log_inference() works with only required fields."""
+        row_id = log_inference(
+            agent_name="min-agent",
+            provider="ollama",
+            model="hermes3:8b",
+            endpoint="/v1/chat/completions",
+        )
+        self.assertIsNotNone(row_id)
+        
+        db = get_db_path()
+        conn = sqlite3.connect(str(db))
+        conn.row_factory = sqlite3.Row
+        cursor = conn.cursor()
+        cursor.execute("SELECT * FROM request_log WHERE id = ?", (row_id,))
+        row = cursor.fetchone()
+        conn.close()
+        
+        self.assertEqual(row["agent_name"], "min-agent")
+        self.assertEqual(row["tokens_in"], None)
+        self.assertEqual(row["latency_ms"], None)
+    
+    def test_log_inference_error_status(self):
+        """log_inference() records error status with message."""
+        row_id = log_inference(
+            agent_name="err-agent",
+            provider="openrouter",
+            model="claude-opus-4-6",
+            endpoint="/v1/chat/completions",
+            status="error",
+            error_message="Rate limit exceeded",
+        )
+        self.assertIsNotNone(row_id)
+        
+        db = get_db_path()
+        conn = sqlite3.connect(str(db))
+        cursor = conn.cursor()
+        cursor.execute("SELECT error_message FROM request_log WHERE id = ?", (row_id,))
+        row = cursor.fetchone()
+        conn.close()
+        
+        self.assertEqual(row[0], "Rate limit exceeded")
+    
+    def test_query_requests_filters_by_agent(self):
+        """query_requests() filters by agent_name."""
+        log_inference(agent_name="agent-a", provider="p1", model="m1", endpoint="/e1")
+        log_inference(agent_name="agent-a", provider="p1", model="m1", endpoint="/e2")
+        log_inference(agent_name="agent-b", provider="p1", model="m1", endpoint="/e3")
+        
+        rows = query_requests(agent_name="agent-a")
+        self.assertEqual(len(rows), 2)
+        for row in rows:
+            self.assertEqual(row["agent_name"], "agent-a")
+    
+    def test_query_requests_filters_by_provider(self):
+        """query_requests() filters by provider."""
+        log_inference(agent_name="a1", provider="anthropic", model="claude-4", endpoint="/e")
+        log_inference(agent_name="a2", provider="openrouter", model="claude-4", endpoint="/e")
+        log_inference(agent_name="a3", provider="anthropic", model="claude-4", endpoint="/e")
+        
+        rows = query_requests(provider="anthropic")
+        self.assertEqual(len(rows), 2)
+        for row in rows:
+            self.assertEqual(row["provider"], "anthropic")
+    
+    def test_query_requests_time_window(self):
+        """query_requests() respects hours parameter."""
+        log_inference(agent_name="time-agent", provider="test", model="m", endpoint="/e")
+        
+        rows = query_requests(hours=1)
+        self.assertGreaterEqual(len(rows), 1)
+        
+        # 24-hour window should include at least what 1-hour includes
+        rows_recent = query_requests(hours=24)
+        self.assertGreaterEqual(len(rows_recent), len(rows))
+    
+    def test_did_agent_call_provider_positive(self):
+        """did_agent_call_provider() returns True when agent called provider."""
+        log_inference(
+            agent_name="codex-agent",
+            provider="anthropic",
+            model="claude-sonnet-4-20250514",
+            endpoint="/v1/messages",
+            status="success",
+        )
+        
+        result = did_agent_call_provider(
+            agent_name="codex-agent",
+            provider="anthropic",
+            hours=24,
+        )
+        self.assertTrue(result)
+    
+    def test_did_agent_call_provider_negative_wrong_agent(self):
+        """did_agent_call_provider() returns False for non-matching agent."""
+        log_inference(
+            agent_name="other-agent",
+            provider="anthropic",
+            model="claude-sonnet-4-20250514",
+            endpoint="/v1/messages",
+        )
+        
+        result = did_agent_call_provider(
+            agent_name="codex-agent",
+            provider="anthropic",
+        )
+        self.assertFalse(result)
+    
+    def test_did_agent_call_provider_negative_wrong_provider(self):
+        """did_agent_call_provider() returns False for non-matching provider."""
+        log_inference(
+            agent_name="codex-agent",
+            provider="ollama",
+            model="hermes3:8b",
+            endpoint="/v1/chat/completions",
+        )
+        
+        result = did_agent_call_provider(
+            agent_name="codex-agent",
+            provider="anthropic",
+        )
+        self.assertFalse(result)
+    
+    def test_did_agent_call_provider_min_success_count(self):
+        """did_agent_call_provider() respects min_success_count."""
+        log_inference(
+            agent_name="agent-x",
+            provider="p",
+            model="m",
+            endpoint="/e",
+            status="success",
+        )
+        
+        self.assertTrue(did_agent_call_provider("agent-x", "p", min_success_count=1))
+        self.assertFalse(did_agent_call_provider("agent-x", "p", min_success_count=2))
+    
+    def test_log_and_query_by_status(self):
+        """query_requests() can filter by status."""
+        log_inference(agent_name="a", provider="p", model="m", endpoint="/e", status="success")
+        log_inference(agent_name="a", provider="p", model="m", endpoint="/e", status="error")
+        log_inference(agent_name="a", provider="p", model="m", endpoint="/e", status="timeout")
+        
+        success_rows = query_requests(status="success")
+        error_rows = query_requests(status="error")
+        
+        self.assertEqual(len(success_rows), 1)
+        self.assertEqual(len(error_rows), 1)
+    
+    def test_get_recent_activity_summary(self):
+        """get_recent_activity_summary() returns structured data."""
+        log_inference(agent_name="agent-1", provider="anthropic", model="claude-4", endpoint="/e", latency_ms=1000)
+        log_inference(agent_name="agent-1", provider="anthropic", model="claude-4", endpoint="/e", latency_ms=2000)
+        
+        summary = get_recent_activity_summary(hours=24)
+        self.assertIn("by_agent_provider", summary)
+        found = any(
+            r.get("agent_name") == "agent-1" and r.get("provider") == "anthropic"
+            for r in summary.get("by_agent_provider", [])
+        )
+        self.assertTrue(found, "Should find agent-1/anthropic in summary")
+
+
+if __name__ == "__main__":
+    unittest.main()