Sensever bots: features and limits
A reference for anyone designing a bot: how much is allowed, what the bot sees, and how to protect the bot and people. All numbers are current platform limits.
Owner and bots
What | Limit |
|---|---|
Who can create a bot | A person with a Sensever account and a verified email or phone number; otherwise |
How many bots | Up to 20 active bots per person; deleted bots do not count. Beyond that: |
Username | 5–32 characters: a Latin letter first, then Latin letters, digits and single |
Username availability | A username is unique across bots, people and channel addresses. A deleted bot's username is not released. The reserved names |
The | Bots only: people and channels cannot take such a name ( |
Bot name | 1–128 characters |
Token |
|
Webhook secret | 64 hexadecimal characters. Changes together with the token |
If the owner's account is blocked, the bot stops working: the API does not respond, updates are not delivered and expire after 24 hours.
Updates
What | Limit |
|---|---|
Update retention | 24 hours; then |
Delivery log | The last 100 updates; completed ones are deleted after 7 days |
|
|
Webhook | HTTPS, port 443, a domain name, a public address, no redirects |
Webhook response time | 2 seconds to connect, 5 seconds to respond |
Update size | Up to 256 KB |
Delivery retries | 5 s, 10 s, 20 s … at least once every 10 minutes until the update expires |
| Any of the 11 update types; an empty list means all |
Messages
What | Limit |
|---|---|
Text | Up to 16,384 characters |
Files per message | Up to 10 |
File size | Up to 2 GB; over 50 MB only via tus upload |
File from a URL |
|
File in a Telegram-compatible request | Up to 20 MB |
File links in a message | Temporary, valid for minutes; read the message again to get new ones |
Mentions | Up to 100 per message |
Reaction | Up to 64 characters; up to 12 reactions per member on a message; up to 32 distinct reactions per message |
Poll | Question up to 500 characters, 2–10 options of up to 200 characters, a single answer |
Request body | Up to 256 KB of JSON |
Subscribers by email or phone | Up to 1,000 new addresses per 10 minutes per bot (an address already looked up is remembered for 10 minutes and does not count), and up to 300 "not a subscriber" answers per hour. Beyond that: |
Buttons, commands, profile
What | Limit |
|---|---|
Inline buttons | Up to 8 rows of 8 buttons; label up to 64 characters; |
Answer to a tap | Once, within 15 minutes; text up to 200 characters. The person waits up to 10 seconds for the answer |
Keyboard below the input field | Up to 12 rows of 8 buttons; label up to 64 characters; placeholder up to 64 characters |
Commands | Up to 100 in a list; a command is 1–32 characters of |
Description | Up to 512 characters |
Short description | Up to 120 characters |
Avatar | JPEG, PNG or WebP, up to 10 MB and up to 20 megapixels |
| 1–64 characters of |
Inline query | Up to 256 characters; answer with up to 50 results, once, within 15 minutes; the person waits up to 8 seconds |
Inline query result |
|
Inline mode placeholder | Up to 64 characters |
Rate limits
Limits are counted per bot per minute unless stated otherwise.
What | Limit |
|---|---|
All API requests | 600 per minute |
Sending messages | 120 per minute and 60 per minute per conversation |
Editing and deleting messages | 120 per minute and 60 per minute per conversation |
Uploading files | 120 per minute and 60 per minute per conversation |
| 120 per minute |
| 20 per minute |
"Typing…" | 120 per minute; no more than 2 times per 2 seconds per conversation |
Answers to taps and inline queries | 120 per minute each |
Changing commands and descriptions | 30 per minute each |
File from a URL | 30 per minute |
Telegram-compatible adapter | An additional 1200 requests per minute from one IP address; file downloads, 600 per minute |
When a limit is exceeded, you get 429. How to find out how long to wait is described in "Bot API: basics", section "Exceeding limits".
What the bot sees
Direct chat: everything the person writes, their edits, reactions, taps and votes.
Project request: the visitor's messages until an agent takes the request. The visitor is anonymous:
tu1.…instead of an account number, no avatar. Once an agent has taken the request, the bot no longer reads or writes in it.Group: in privacy mode, only messages addressed to the bot; without it, all messages from people. The bot reads group history as a member, from the moment it joined, regardless of privacy mode.
Channel: all posts.
Inline mode: only the query text, the author and the chat type, but not the chat itself.
Phone number: only if the person tapped the
request_contactbutton and confirmed sending; the number is verified in the Sensever account.
The bot does not see: people's chat lists and contacts, conversations it is not a member of, messages hidden by an access policy, deleted messages. Permissions are checked when each update is delivered and on every request: if the bot was removed, updates about past messages are skipped.
What the bot cannot do
Start a conversation: write to a person who has not opened a chat with the bot, or create a group.
Edit other people's messages — even as an admin.
Vote in polls, bookmark messages or hide messages.
Appoint admins or change the owner's role.
Manage a project, widget settings or other bots.
Reply in a request an agent has taken.
Security
For the bot server
Keep the token and webhook secret on the server only. Do not put them in page code, mobile apps or Mini Apps. If they leak, issue a new token — the webhook secret changes with it.
Verify the webhook signature against the raw body bytes and reject requests whose
X-Sensever-Timestampis more than ±5 minutes off. See the example in "Bot API: receiving updates".Store updates with the unique key
(bot_id, update_id): delivery may repeat.Send messages with a
client_message_idthat stays the same on retry, so a network failure does not create a duplicate.Verify Mini App data on the server: the
hashsignature and the freshness ofauth_date.Store identifiers as strings: they do not fit into JavaScript numbers.
Do not work around refusals. A
409 conversation_owned,403or401response is a signal to stop, not to retry the request another way.
For people
The bot sees only what a member in its place would see, and only while it is a member.
In groups with a bot that has privacy mode off, members see the label "Bot · sees all messages".
A message sent through inline mode is marked "via @bot"; the marker cannot be forged or removed.
A person can block the bot: the bot stops receiving updates from this person and cannot write to them (
403 blocked).
Comments