diff --git a/src/bot.py b/src/bot.py
index 3ba3551..439b65a 100644
--- a/src/bot.py
+++ b/src/bot.py
@@ -15,6 +15,7 @@
 from config import config
 from google_auth import generate_auth_url, TokenStorage
 from timebox import build_timebox_handler
+from search import search_command
 import drive_handler
 
 # Configure logging
@@ -38,6 +39,7 @@ async def start_command(update: Update, context: ContextTypes.DEFAULT_TYPE) -> N
             "You are authenticated with Google Drive.\n"
             "Send me any message and I'll save it to your Drive.\n\n"
             "Commands:\n"
+            "/search - Search your captured notes\n"
             "/timebox - Plan a timeboxed schedule for tomorrow\n"
             "/status - Check connection status\n"
             "/logout - Disconnect Google Drive\n"
@@ -69,6 +71,7 @@ async def help_command(update: Update, context: ContextTypes.DEFAULT_TYPE) -> No
         "Edit a message here and it updates in Drive too.\n\n"
         "Commands:\n"
         "/authenticate - Connect your Google Drive\n"
+        "/search - Search your captured notes\n"
         "/timebox - Plan a timeboxed schedule for tomorrow\n"
         "/status - Check connection status\n"
         "/logout - Disconnect Google Drive\n"
@@ -309,6 +312,7 @@ def register_handlers(application: Application) -> None:
     application.add_handler(CommandHandler("authenticate", authenticate_command))
     application.add_handler(CommandHandler("status", status_command))
     application.add_handler(CommandHandler("logout", logout_command))
+    application.add_handler(CommandHandler("search", search_command))
 
     application.add_handler(build_timebox_handler())
 
diff --git a/src/drive_handler.py b/src/drive_handler.py
index bc55924..6f88518 100644
--- a/src/drive_handler.py
+++ b/src/drive_handler.py
@@ -250,6 +250,26 @@ def _replace_message_content(
     return file_content[:header_start] + new_block + file_content[content_end:]
 
 
+def list_markdown_files(service, folder_id: str) -> list[dict]:
+    """Return the markdown files in the folder as [{id, name}, ...] (read-only)."""
+    try:
+        results = service.files().list(
+            q=f"mimeType='{MARKDOWN_MIME_TYPE}' and trashed=false and parents='{folder_id}'",
+            spaces='drive',
+            fields='files(id, name)',
+            pageSize=1000,
+        ).execute()
+        return results.get('files', [])
+    except Exception as e:
+        logger.error(f"Failed to list markdown files: {e}")
+        return []
+
+
+def read_file(service, file_id: str) -> str | None:
+    """Public read of a Drive file's text content, or None on failure."""
+    return _download_file_content(service, file_id)
+
+
 def _download_file_content(service, file_id: str) -> str | None:
     """Download a file's content from Drive."""
     try:
diff --git a/src/search.py b/src/search.py
new file mode 100644
index 0000000..3ef2a4b
--- /dev/null
+++ b/src/search.py
@@ -0,0 +1,107 @@
+"""/search: keyword search across the captured daily notes in Drive.
+
+Thin Telegram glue. The matching/parsing helpers are pure functions so they
+can be unit-tested without Drive. Scans the most recent daily files only, so a
+long history stays cheap.
+"""
+
+import logging
+import re
+
+from telegram import Update
+from telegram.ext import ContextTypes
+
+from config import config
+import drive_handler
+
+logger = logging.getLogger(__name__)
+
+_BLOCK_RE = re.compile(r"<!-- msg_id: \d+ -->\n")
+_DAILY_RE = re.compile(r"^(\d{4}-\d{2}-\d{2})\.md$")
+
+_MAX_FILES = 60      # how many recent daily files to scan
+_MAX_HITS = 20       # how many matches to show
+_SNIPPET_LEN = 160   # per-hit snippet length
+_MAX_REPLY = 3900
+
+
+def parse_blocks(content: str) -> list[str]:
+    """Captured note texts in order, stripped of the msg_id markers."""
+    parts = _BLOCK_RE.split(content)
+    return [p.strip() for p in parts[1:] if p.strip()]
+
+
+def find_matches(content: str, query: str) -> list[str]:
+    """Notes in `content` that contain `query` (case-insensitive substring)."""
+    q = query.lower()
+    return [n for n in parse_blocks(content) if q in n.lower()]
+
+
+def _snippet(note: str) -> str:
+    """Collapse a note to a single trimmed line for compact display."""
+    one_line = " ".join(note.split())
+    if len(one_line) > _SNIPPET_LEN:
+        one_line = one_line[:_SNIPPET_LEN].rstrip() + "…"
+    return one_line
+
+
+def _date_of(file_name: str) -> str:
+    """The YYYY-MM-DD date encoded in a daily file name, or the raw name."""
+    m = _DAILY_RE.match(file_name)
+    return m.group(1) if m else file_name
+
+
+async def search_command(update: Update, context: ContextTypes.DEFAULT_TYPE) -> None:
+    """Handle /search <query> - find captured notes matching the query."""
+    token_storage = context.bot_data.get("token_storage")
+    user_id = update.effective_user.id
+    if not token_storage or not token_storage.is_authenticated(user_id):
+        await update.message.reply_text(
+            "Please authenticate with Google Drive first using /authenticate"
+        )
+        return
+
+    query = " ".join(context.args).strip() if context.args else ""
+    if not query:
+        await update.message.reply_text("Usage: /search <query>\ne.g. /search dentist")
+        return
+
+    service = drive_handler.get_drive_service(user_id, token_storage)
+    if not service:
+        await update.message.reply_text(
+            "Your Google Drive session expired — please /logout then /authenticate."
+        )
+        return
+
+    folder_id = drive_handler.get_or_create_folder(service, config.drive_folder_name)
+    if not folder_id:
+        await update.message.reply_text(
+            "Couldn't reach your Drive folder — please try again later."
+        )
+        return
+
+    files = drive_handler.list_markdown_files(service, folder_id)
+    # Newest day first; cap the scan so a long history stays cheap.
+    files = sorted(files, key=lambda f: f.get("name", ""), reverse=True)[:_MAX_FILES]
+
+    hits: list[tuple[str, str]] = []  # (date, snippet)
+    for f in files:
+        if len(hits) >= _MAX_HITS:
+            break
+        content = drive_handler.read_file(service, f["id"]) or ""
+        for note in find_matches(content, query):
+            hits.append((_date_of(f.get("name", "")), _snippet(note)))
+            if len(hits) >= _MAX_HITS:
+                break
+
+    if not hits:
+        await update.message.reply_text(f'No notes found for "{query}".')
+        return
+
+    lines = [f'🔎 {len(hits)} match(es) for "{query}":', ""]
+    for d, snip in hits:
+        lines.append(f"[{d}] {snip}")
+    reply = "\n".join(lines)
+    if len(reply) > _MAX_REPLY:
+        reply = reply[:_MAX_REPLY] + "\n…(truncated)"
+    await update.message.reply_text(reply)
diff --git a/tests/test_search.py b/tests/test_search.py
new file mode 100644
index 0000000..e4aca3c
--- /dev/null
+++ b/tests/test_search.py
@@ -0,0 +1,37 @@
+from search import parse_blocks, find_matches, _snippet, _date_of
+
+CONTENT = (
+    "# Telegram Messages\n\n"
+    "<!-- msg_id: 1 -->\nBook dentist appointment\n"
+    "<!-- msg_id: 2 -->\nGroceries: milk, eggs\n"
+    "<!-- msg_id: 3 -->\nDENTIST follow-up next month\n"
+)
+
+
+def test_parse_blocks():
+    assert parse_blocks(CONTENT) == [
+        "Book dentist appointment",
+        "Groceries: milk, eggs",
+        "DENTIST follow-up next month",
+    ]
+
+
+def test_find_matches_is_case_insensitive():
+    hits = find_matches(CONTENT, "dentist")
+    assert hits == ["Book dentist appointment", "DENTIST follow-up next month"]
+
+
+def test_find_matches_none():
+    assert find_matches(CONTENT, "passport") == []
+
+
+def test_snippet_collapses_and_truncates():
+    assert _snippet("line one\n  line  two") == "line one line two"
+    long = "x" * 300
+    out = _snippet(long)
+    assert out.endswith("…") and len(out) <= 161
+
+
+def test_date_of():
+    assert _date_of("2026-06-25.md") == "2026-06-25"
+    assert _date_of("weird.md") == "weird.md"