diff --git a/baseline/baseline.md b/baseline/baseline.md
index b350783..578938a 100644
--- a/baseline/baseline.md
+++ b/baseline/baseline.md
@@ -37,10 +37,17 @@ nohup gzip -cd /mnt/langsplit/2015_32_kv.gz | ~/DataCollection/baseline/langstat
 If you are collecting data for a language direction for which you already earlier collected data from the reverse direction, please see an optimized process in the appendix.
 
 ## Step 3: Look up where these URLs appear in CommonCrawl S3
+
+### Option 1 (if you have build your own location database)
 ```
 nohup cat candidates.en-de | nice ~/DataCollection/baseline/locate_candidates.py - - -server='http://statmt.org:8084/query_prefix' > candidates.en-de.locations 2> locate.log &
 ```
 
+### Option 2 (use the [CommonCrawl Index API](http://commoncrawl.org/2015/04/announcing-the-common-crawl-index/))
+```
+nohup cat candidates.en-de | nice ~/DataCollection/baseline/locate_candidates_cc_index_api.py - - > candidates.en-de.locations 2> locate.log &
+```
+
 ## Step 4: Download pages from CommonCrawl S3 and extract text
 For certain language pairs we provide the `.locations` files in compressed form in our releases on https://github.com/ModernMT/DataCollection/releases. You can use these files to start the process in this step.
 ```
diff --git a/baseline/locate_candidates_cc_index_api.py b/baseline/locate_candidates_cc_index_api.py
new file mode 100755
index 0000000..d5c3391
--- /dev/null
+++ b/baseline/locate_candidates_cc_index_api.py
@@ -0,0 +1,105 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+import sys
+import json
+import tldextract
+import re
+import requests
+import urllib
+
+COMMONCRAWL_S3_URL = "https://commoncrawl.s3.amazonaws.com"
+COMMONCRAWL_INDEX_URL = "http://index.commoncrawl.org"
+
+INVALID_URL = "123"
+INVALID_CRAWL = "abc"
+
+def make_full_filename(filepath):
+    return '/'.join([COMMONCRAWL_S3_URL, filepath])
+
+def make_query_url(crawl, url):
+    params = {
+        "base_url": COMMONCRAWL_INDEX_URL,
+        "crawl_id": crawl.replace('_', '-'),
+        "url": urllib.quote(url, safe='')      # Percent encode URL.
+    }
+
+    query = "{base_url}/CC-MAIN-{crawl_id}-index?url={url}&output=json&limit=1"
+    return query.format(**params)
+
+def get_location(session, url, crawl):
+    """ Returns success and location """
+    query_url = make_query_url(crawl, url)
+    try:
+        r = session.get(query_url)
+        result = r.json()
+    except:
+        return False, None
+
+    try:
+        data = {
+            "filename": make_full_filename(result["filename"]),
+            "length": result["length"],
+            "mime": result["mime"],
+            "offset": result["offset"]
+        }
+    except KeyError:
+        return False, None
+
+    return True, data
+
+
+def report_error(url, crawl, errors, total):
+    percentage = 100. * errors / total
+    sys.stderr.write("Errors: %d/%d = %.2f%%\t%s\t%s\n" %
+                     (errors, total, percentage, crawl, url))
+
+
+if __name__ == "__main__":
+    import argparse
+    parser = argparse.ArgumentParser()
+    parser.add_argument('candidates', type=argparse.FileType('r'),
+                        help='file containing candidates')
+    parser.add_argument('outfile', type=argparse.FileType('w'),
+                        default=sys.stdout)
+    parser.add_argument('-kv', help='input is a .kv.gz file',
+                        default=False, action="store_true")
+    args = parser.parse_args(sys.argv[1:])
+
+    total_lines, total_errors = 0, 0
+    with requests.Session() as session:
+        for line in args.candidates:
+            total_lines += 1
+            line = line.decode("utf-8")
+            if args.kv:
+                # Lines have the format:
+                # {domain} {url} {crawl}\t{language_data}
+                url_data, _ = line.strip().split('\t')
+                _, src_url, src_crawl = url_data.strip().split()
+                tgt_success = False
+            else:
+                # Lines have the format:
+                # {stripped_url} {src_url} {src_crawl} {tgt_url} {tgt_crawl}
+                _, src_url, src_crawl, tgt_url, tgt_crawl = line.strip().split()
+
+            src_success, src_loc = get_location(session, src_url, src_crawl)
+            if not src_success:
+                total_errors += 1
+                report_error(src_url, src_crawl, total_errors, total_lines)
+
+            if not args.kv:
+                tgt_success, tgt_loc = get_location(session, tgt_url, tgt_crawl)
+                if not tgt_success:
+                    total_errors += 1
+                    report_error(tgt_url, tgt_crawl, total_errors, total_lines)
+
+            if src_success and tgt_success:
+                args.outfile.write("%s\t%s\t%s\n" %
+                                   (src_url, src_crawl, json.dumps(src_loc)))
+                args.outfile.write("%s\t%s\t%s\n" %
+                                   (tgt_url, tgt_crawl, json.dumps(tgt_loc)))
+            elif args.kv and src_success:
+                args.outfile.write("%s\t%s\t%s\n" %
+                                   (src_url, src_crawl, json.dumps(src_loc)))
+
+    sys.stderr.write("Done: ")
+    report_error(tgt_url, tgt_crawl, total_errors, total_lines)