diff --git a/digest-bot.py b/digest-bot.py
new file mode 100644
index 0000000..e0848d2
--- /dev/null
+++ b/digest-bot.py
@@ -0,0 +1,115 @@
+#!/usr/bin/env python3
+import csv
+import requests
+import os
+import sys
+import json
+
+# Config
+CSV_PATH = '/app/data/posts-latest.csv'
+CLAUDE_API = 'https://api.anthropic.com/v1/messages'
+CLAUDE_KEY = os.getenv('ANTHROPIC_API_KEY')
+MASTODON_URL = 'https://zpravobot.news'
+
+# Bot selection
+bot_name = sys.argv[1] if len(sys.argv) > 1 else 'zpravobot'
+TOKENS = {
+    'zpravobot': os.getenv('TOKEN_ZPRAVOBOT'),
+    'pozitivni': os.getenv('TOKEN_POZITIVNI'),
+    'sarkasticky': os.getenv('TOKEN_SARKASTICKY')
+}
+
+if bot_name not in TOKENS:
+    print(f"❌ Unknown bot: {bot_name}")
+    sys.exit(1)
+
+TOKEN = TOKENS[bot_name]
+
+if not TOKEN or not CLAUDE_KEY:
+    print(f"❌ Missing env variables")
+    sys.exit(1)
+
+# 1. Load posts
+try:
+    with open(CSV_PATH, 'r', encoding='utf-8') as f:
+        posts = list(csv.DictReader(f))
+    print(f"📊 Loaded {len(posts)} posts for @{bot_name}")
+except Exception as e:
+    print(f"❌ Error loading CSV: {e}")
+    sys.exit(1)
+
+# 2. Prepare data for Claude
+posts_sample = posts[:500]  # Limit to 500 posts
+posts_json = json.dumps(posts_sample, ensure_ascii=False)
+
+# 3. Claude API
+print("🤖 Calling Claude API...")
+try:
+    response = requests.post(
+        CLAUDE_API,
+        headers={
+            'x-api-key': CLAUDE_KEY,
+            'anthropic-version': '2023-06-01',
+            'content-type': 'application/json'
+        },
+        json={
+            'model': 'claude-sonnet-4-20250514',
+            'max_tokens': 4000,
+            'messages': [{
+                'role': 'user',
+                'content': f'Vytvoř denní digest pro bot @{bot_name}. Data: {posts_json[:10000]}'
+            }]
+        },
+        timeout=60
+    )
+    
+    if response.status_code != 200:
+        print(f"❌ Claude API error: {response.text}")
+        sys.exit(1)
+    
+    digest = response.json()['content'][0]['text']
+    print(f"✅ Claude response: {len(digest)} chars")
+except Exception as e:
+    print(f"❌ Claude API exception: {e}")
+    sys.exit(1)
+
+# 4. Split to 2 toots (500 chars limit)
+toot1 = digest[:500]
+toot2 = digest[500:1000] if len(digest) > 500 else None
+
+# 5. Publish toot 1
+print("📤 Publishing toot 1...")
+try:
+    r1 = requests.post(
+        f'{MASTODON_URL}/api/v1/statuses',
+        headers={'Authorization': f'Bearer {TOKEN}'},
+        json={'status': toot1}
+    )
+    
+    if r1.status_code not in [200, 201]:
+        print(f"❌ Mastodon error: {r1.text}")
+        sys.exit(1)
+    
+    toot1_id = r1.json()['id']
+    print(f"✅ Toot 1 published: {toot1_id}")
+except Exception as e:
+    print(f"❌ Mastodon exception: {e}")
+    sys.exit(1)
+
+# 6. Publish toot 2 (if exists)
+if toot2:
+    print("📤 Publishing toot 2...")
+    try:
+        r2 = requests.post(
+            f'{MASTODON_URL}/api/v1/statuses',
+            headers={'Authorization': f'Bearer {TOKEN}'},
+            json={
+                'status': toot2,
+                'in_reply_to_id': toot1_id
+            }
+        )
+        print(f"✅ Toot 2 published (thread)")
+    except Exception as e:
+        print(f"⚠️ Toot 2 failed: {e}")
+
+print(f"✅ Done! Published to @{bot_name}")