diff --git a/TODO b/TODO
index 4956738..7f46939 100644
--- a/TODO
+++ b/TODO
@@ -5,10 +5,11 @@ TODO list
 - find (authors) duplicates
 + remove command
 - stats command
-- add query support to list command (cf beets)
++- add query support to list command (cf beets)
 - FIX open on ubuntu
-- edit bibtex on add
++- edit bibtex on add
 - option to add tag on add
 - paths in metadata must be relative
 - include package data in setup.py (see setup.py for more info)
 - in the script papers, why do we use a dictionary and not a list? it may be confusing, the key is not the subparser name. The subparser name is defined in the cmd files
+- command to output a bibfile from several citekey or even a research results
diff --git a/papers/commands/__init__.py b/papers/commands/__init__.py
index 074dff3..8d013b5 100644
--- a/papers/commands/__init__.py
+++ b/papers/commands/__init__.py
@@ -10,3 +10,4 @@ import remove_cmd
 import websearch_cmd
 
 import devadd_cmd
+import devlist_cmd
diff --git a/papers/commands/devlist_cmd.py b/papers/commands/devlist_cmd.py
new file mode 100644
index 0000000..2841877
--- /dev/null
+++ b/papers/commands/devlist_cmd.py
@@ -0,0 +1,53 @@
+from .. import pretty
+from .. import repo
+
+#TODO how to have an empty case, i.e. cmd is not provided -> list all
+#TODO add formatting option so that the output can be piped
+
+def parser(subparsers, config):
+    parser = subparsers.add_parser('devlist', help="list papers")
+    parser.add_argument('cmd', help='experimental option "year: 2000 or labels: bite"')
+    return parser
+
+
+
+def command(config, ui, cmd):
+    rp = repo.Repository.from_directory(config)
+
+    tests = cmd.split()
+
+    articles = []
+    for n, p in enumerate(rp.all_papers()):
+        if test_paper(tests, p):
+            bibdesc = pretty.bib_oneliner(p.bibentry, color=ui.color)
+            articles.append((u'{num:d}: [{citekey}] {descr} {labels}'.format(
+                num=int(n),
+                citekey=ui.colored(rp.citekeys[n], 'purple'),
+                descr=bibdesc,
+                labels=ui.colored(''.join(p.metadata.get('labels', [])), 'purple'),
+                )).encode('utf-8'))
+    ui.print_('\n'.join(articles))
+
+#TODO author is not implemented, should we do it by last name only or more complex
+#TODO implement search by type of document
+def test_paper(tests, p):
+    for test in tests:
+        tmp = test.split(':')
+        if tmp[0] == 'all':
+            return True
+        elif len(tmp) != 2:
+            raise ValueError('command not valid')
+
+        field = tmp[0]
+        value = tmp[1]
+
+        if field == 'labels':
+            if value not in p.metadata['labels']:
+                return False
+        else:
+            if p.bibentry.fields.has_key(field):
+                if value not in p.bibentry.fields[field]:
+                    return False
+            else:
+                return False
+    return True
diff --git a/papers/papers b/papers/papers
index cc1fe38..d7c08a5 100755
--- a/papers/papers
+++ b/papers/papers
@@ -23,6 +23,7 @@ cmds = collections.OrderedDict([
         ])
 
 cmds.update(collections.OrderedDict([('devadd', commands.devadd_cmd)]))
+cmds.update(collections.OrderedDict([('devlist', commands.devlist_cmd)]))
 
 config = configs.read_config()
 ui = UI(config)