You need to use the execute method

The documentation for this method is very clear that this method allows you to perform up to 25 requests to the API in 1 request to the method.

As far as I understand you, you need to collect all comments from the post (if not, write in the comments).
But here you need to immediately put such restrictions:
Because requests are only 25 , therefore we can get 25 * 100 comments. ( 100 because for 1 request to wall.getComments we will receive no more than 100 comments )

The next item is to make a code that will get these 2500 comments from a certain post.

 var post_id = Args.post_id; var owner_id = Args.owner_id; var count = Args.count; var offset = Args.offset; if(post_id == null || owner_id == null){ return {"count": 0, "items": []}; // Если параметры не заданы, возвращаем пустой массив } if(count == null){ count = 10; // Значение по умолчанию } if(offset == null){ offset = 0; // Значение по умолчанию } var i = 0; var items = []; while(i < 25 && count > items.length){ var cur_count = count - items.length; if(cur_count > 100){ cur_count = 100; } items = items + API.wall.getComments({"owner_id": owner_id, "post_id": post_id, "offset": offset+100*i, "count": cur_count})["items"]; i = i + 1; } return {"count": items.length, "items": items}; 

This code can be passed as an argument to execute (the code argument), and along with this argument, the post_id , ower_id and others.
The second option is to create a procedure in your application and already call with the token received for your application to call this procedure with the parameters owner_id , post_id and the rest.

How to get more than 2500 comments?

To do this, you just need to call this code several times with different parameters of the offset (by constantly increasing it by 2500)
Python3 example :

 from math import ceil import json,time import vk_api vk = vk_api.VkApi(token='...') # Ваш токен api = vk.get_api() posts = api.wall.getById(posts='{owner_id}_{post_id}'.format(owner_id='<owner_id>', post_id='<post_id>')) # Где <owner_id> - id пользователя/группы, post_id - id поста. *см. комментарии после кода if not posts: print('No posts found!') post = posts[0] posts = [] for offset in range(ceil(post['comments']['count'] // 2500)): posts.extend(api.execute.getCommentsFromPost(owner_id='<owner_id>', post_id='<post_id>', offset=2500*offset, count=2500)['items']) # тут всё по аналогии, как и в первый раз time.sleep(0.34) # избегаем flood control with open('output.json', 'w') as file: json.dump(posts, file) 

It should be noted that community id should be specified with a minus, for example -1 .
It is also worth mentioning that by repeatedly invoking the same method, the VC blocks access to it.