diff --git a/docs/sphinx/en/how_to_use/standard_usage.rst b/docs/sphinx/en/how_to_use/standard_usage.rst
index 79d11f8c..49c394a9 100644
--- a/docs/sphinx/en/how_to_use/standard_usage.rst
+++ b/docs/sphinx/en/how_to_use/standard_usage.rst
@@ -10,7 +10,11 @@ Usage of ``tenes_std``
    $ tenes_std std.toml
 
 
-- Takes a file as an argument
+- Takes input files as arguments
+   - Multiple input files can be specified
+   - When parameters are duplicated, ``tenes_std`` stops with an error
+   - Sections that can be specified multiple times, such as ``[[observable.onesite]]``, can be specified in multiple input files simultaneously
+      - In this case, the sections in the latter input file are appended to those in the former input file
 - Output an input file for ``tenes``
 - Command line options are as follows
    - ``--help``
diff --git a/docs/sphinx/ja/how_to_use/standard_usage.rst b/docs/sphinx/ja/how_to_use/standard_usage.rst
index caac478b..03b56812 100644
--- a/docs/sphinx/ja/how_to_use/standard_usage.rst
+++ b/docs/sphinx/ja/how_to_use/standard_usage.rst
@@ -7,10 +7,14 @@
 
 .. code:: bash
 
-   $ tenes_std std.toml
+   $ tenes_std std.toml [std_2.toml ...]
 
 
 - 引数としてファイルを取ります
+   - 複数の入力ファイルからひとつの出力ファイルを生成可能です
+   - 同じ名前のパラメータを複数の入力で指定した場合はエラー終了します
+   - ``[[observable.onesite]]`` などの複数指定できるセクションについては同時に指定可能です
+      - 後ろのファイルの内容が前のファイルの内容に追記されます
 - ``tenes`` の入力ファイルを出力します
 - コマンドラインオプションは以下の通りです
    - ``--help``
diff --git a/sample/05_hardcore_boson_triangular/run.py b/sample/05_hardcore_boson_triangular/run.py
index d9ff6920..ffcb4137 100644
--- a/sample/05_hardcore_boson_triangular/run.py
+++ b/sample/05_hardcore_boson_triangular/run.py
@@ -96,13 +96,10 @@
             cmd = f"tenes_simple {simple_toml} -o {std_toml}"
             subprocess.call(cmd.split())
 
+            cmd = f"tenes_std -o {input_toml} {std_toml}"
             if calculate_sq:
-                # append twosite observable <n_i n_j> to std.toml
-                with open(std_toml, "a") as f:
-                    with open("nn_obs.toml") as fin:
-                        for line in fin:
-                            f.write(line)
-            cmd = f"tenes_std {std_toml} -o {input_toml}"
+                # twosite observable <n_i n_j> are also calculated
+                cmd += " nn_obs.toml"
             subprocess.call(cmd.split())
 
             cmd = f"{MPI_cmd} tenes {input_toml}"
diff --git a/test/data/output_std_mode.toml b/test/data/output_std_mode.toml
index 35717d20..ad3c2064 100644
--- a/test/data/output_std_mode.toml
+++ b/test/data/output_std_mode.toml
@@ -56,6 +56,14 @@ bonds = """
 1 1 1
 """
 ops = [0, 1]
+[[observable.multisite]]
+name = "multisite"
+group = 0
+dim= [2, 2, 2]
+multisites = """
+0 1 0 0 1
+"""
+ops = [0, 0, 0]
 
 [evolution]
 [[evolution.simple]]
diff --git a/test/std_mode.py.in b/test/std_mode.py.in
index a0d9a1f9..fedd809b 100644
--- a/test/std_mode.py.in
+++ b/test/std_mode.py.in
@@ -44,6 +44,7 @@ cmd = [
     "-o",
     join("output_std_mode.toml"),
     join("data", "std_mode.toml"),
+    join("data", "std_mode_multi.toml"),
 ]
 subprocess.call(cmd)
 
@@ -101,6 +102,28 @@ if not obs_result:
     print('check for the section "observable.twosite" fails')
     result = False
 
+obs_result = True
+if len(res['observable']['multisite']) != len(ref['observable']['multisite']):
+    obs_result = False
+    result = False
+else:
+    for obs_res, obs_ref in zip(res['observable']['multisite'], ref['observable']['multisite']):
+        obs_result = obs_result and obs_res['name'] == obs_ref['name']
+        obs_result = obs_result and obs_res['group'] == obs_ref['group']
+        obs_result = obs_result and obs_res['multisites'] == obs_ref['multisites']
+        if 'elements' in obs_ref:
+            obs_result = obs_result and 'elements' in obs_res
+            obs_result = obs_result and obs_res['dim'] == obs_ref['dim']
+            elem_res = load_str_as_array(obs_res['elements'])
+            elem_ref = load_str_as_array(obs_ref['elements'])
+            obs_result = obs_result and np.allclose(elem_ref, elem_res)
+        else:
+            obs_result = obs_result and 'elements' not in obs_res
+            obs_result = obs_result and obs_res['ops'] == obs_ref['ops']
+if not obs_result:
+    print('check for the section "observable.multisite" fails')
+    result = False
+
 
 for name in ('simple', 'full'):
     evo_result = True
diff --git a/tool/tenes_std.py b/tool/tenes_std.py
index 240a072d..4eb0c15e 100644
--- a/tool/tenes_std.py
+++ b/tool/tenes_std.py
@@ -14,7 +14,6 @@
 # You should have received a copy of the GNU General Public License
 # along with this program. If not, see http://www.gnu.org/licenses
 
-from collections import namedtuple
 from itertools import product
 from typing import (
     TextIO,
@@ -89,6 +88,75 @@ def value_to_str(v) -> str:
         return "{}".format(v)
 
 
+def merge_input_dict(
+    d1: dict, d2: dict
+) -> None:
+    section1 = d1.get("parameter", {})
+    section2 = d2.get("parameter", {})
+    subsection_names = ("general", "simple_update", "full_update", "ctm", "random")
+    for name in subsection_names:
+        sub1 = section1.get(name, {})
+        sub2 = section2.get(name, {})
+        for k in sub1.keys():
+            if k in sub2:
+                msg = f"parameter.{name}.{k} is defined in multiple input files"
+                raise RuntimeError(msg)
+        for k in sub2.keys():
+            sub1[k] = sub2[k]
+        if len(sub1) > 0:
+            section1[name] = sub1
+    if len(section1) > 0:
+        d1["parameter"] = section1
+
+    section_names = ("correlation", "correlation_length")
+    for section_name in section_names:
+        section1 = d1.get(section_name, {})
+        section2 = d2.get(section_name, {})
+        for name in section1.keys():
+            if name in section2:
+                msg = f"{section_name}.{name} is defined in multiple input files"
+                raise RuntimeError(msg)
+        for name in section2.keys():
+            section1[name] = section2[name]
+        if len(section1) > 0:
+            d1[section_name] = section1
+
+    section1 = d1.get("tensor", {})
+    section2 = d2.get("tensor", {})
+    for k in section1.keys():
+        if k == "unitcell": continue
+        if k in section2:
+            msg = f"tensor.{k} is defined in multiple input files"
+            raise RuntimeError(msg)
+    for k in section2.keys():
+        if k == "unitcell": continue
+        section1[k] = section2[k]
+    if "unitcell" not in section1:
+        section1["unitcell"] = []
+    for u2 in section2.get("unitcell", []):
+        section1["unitcell"].append(u2)
+    if len(section1) > 0:
+        d1["tensor"] = section1
+
+    section1 = d1.get("hamiltonian", [])
+    section2 = d2.get("hamiltonian", [])
+    for h2 in section2:
+        section1.append(h2)
+    if len(section1) > 0:
+        d1["hamiltonian"] = section1
+
+    section1 = d1.get("observable", {})
+    section2 = d2.get("observable", {})
+    for name in ("onesite", "twosite", "multisite"):
+        sub1 = section1.get(name, [])
+        sub2 = section2.get(name, [])
+        for o2 in sub2:
+            sub1.append(o2)
+        if len(sub1) > 0:
+            section1[name] = sub1
+    if len(section1) > 0:
+        d1["observable"] = section1
+
 class Bond:
     source_site: int
     dx: int
@@ -1172,7 +1240,7 @@ def to_toml(self, f: TextIO):
         description="Input converter for TeNeS", add_help=True
     )
 
-    parser.add_argument("input", help="Input TOML file")
+    parser.add_argument("input", nargs="+", help="Input TOML file")
 
     parser.add_argument(
         "-o", "--output", dest="output", default="input.toml", help="Output TOML file"
@@ -1183,11 +1251,14 @@ def to_toml(self, f: TextIO):
 
     args = parser.parse_args()
 
-    if args.input == args.output:
+    if args.output in args.input:
         print("The names of input and output are the same")
         sys.exit(1)
 
-    param = toml.load(args.input)
+    params = [lower_dict(toml.load(f)) for f in args.input]
+    param = params[0]
+    for p in params[1:]:
+        merge_input_dict(param, p)
     model = Model(param)
 
     with open(args.output, "w") as f: