Chatting with kesarAI
kesarAI answers in any channel it can see. This page covers starting and continuing a chat, the words it understands, what it can look at, and how its models, personality and battery work.
Starting a chat
- Start a message with
ksr(ksr what's the capital of peru), @mention kesarAI, or reply to one of its answers. /ai askworks too; turn on private reply and only you see the answer.- Everyone has their own conversation with kesarAI in each channel. It resets after 30 minutes of quiet.
Replies and joining in
- Reply to kesarAI's answer to you and it continues your conversation — no
ksrneeded. - If someone else replies to your answer without
ksr, kesarAI stays out of it, so people can talk about an answer without setting it off. - To join in, reply to any kesarAI answer with
ksr(or an @mention). kesarAI answers you in your own conversation and sees the answer you replied to. - Reply to a person's message with just
ksror an @mention and kesarAI reacts to that message — settle an argument, explain a joke, fact-check a claim. This works on forwarded messages too.
Words kesarAI understands
| You type | What happens |
|---|---|
ksr think <question> | A slower, deeper answer from the model one step up from the one you chat with (Gemini 3.8 Flash, or Claude Sonnet 5.5 if you chat on Claude Haiku), unless you pick a separate think model on your account page. Uses more battery than a normal answer (see what uses battery). |
ksr quota (or ksr battery) | Shows your AI battery. Free, even when it's empty. |
ksr reset (also ksr new topic, ksr new chat) | Starts fresh in this channel only. |
ksr forget / ksr clear / /ai forget | Clears all your conversations with kesarAI, in every channel. |
ksr fact check <claim> | Checks a claim against the web and gives a verdict with sources. |
ksr remember <something about you> | Saves it to your memory (private). Add “publicly” to share it in this server, or “everywhere” for every server. |
ksr create an image of … / ksr edit … (replying to an image) | Image generation and editing, from your daily image allowance (off unless the bot owner gives you one). |
ksr upscale (replying to an image) | Upscales the image. Uses battery. |
Search control
kesarAI decides on its own when a question needs the web — usually only for recent or changing facts. On Gemini models (the default) searches use Google Search. On Claude models they use Brave Search, with Anthropic's web search as a backup when Brave is unavailable. You can steer it from your own message:
- To skip the search, say “don't search”, “no search”, “without searching”, “don't google” or “don't look it up”. kesarAI answers from what it knows and may mention that it could be out of date.
- To make it search, start your message with “search” (
ksr search …), “look it up”, “google it” or “check online”, or end it with “search it”, “look it up”, “google it”, “check online” or “search the web”. It searches before answering, and the search uses battery like any other. - Only your own message counts. Text elsewhere in the channel never switches searching on or off, and a question about searching (“how does google search work”) is just a question.
Models
- Gemini 3.5 Flash-Lite (Google) answers by default. You can pick Claude Haiku 5.5 or another listed model on your account page; your pick applies in every server. Each model card shows how fast it uses your battery compared with a reply on the default model. Switching is instant — changes apply to your next message.
ksr thinkand/ai thinkthink first, on the model one step up from the one you chat with: Gemini 3.5 Flash-Lite thinks on Gemini 3.8 Flash, Claude Haiku 5.5 on Claude Sonnet 5.5. On the same page you can pick a separate think model: Gemini 3.5 Flash-Lite, Claude Haiku 5.5, Gemini 3.8 Flash or Claude Sonnet 5.5.- Long conversations are shortened by Claude Haiku on every model. If the bot owner has turned on low power, some features pause while your battery is low; your model stays the same.
- If Claude is unavailable, kesarAI says so instead of switching models on its own.
- Room view and the tuned personality run on Claude only.
Reading the model cards
- The times are measured by us, not taken from a spec sheet: the median time to a full answer at the settings the bot uses. Treat them as a ranking, not a promise; they move with load.
- Reads images — if a model can't, pasting a screenshot at it just wastes the turn.
- Streams live — the answer is typed out as it's written. A model marked One block sits silent for its full response time and then posts the whole answer at once, which feels much slower than the same number of seconds spent streaming.
- Want a model that isn't listed? Ask in the support server and we'll try to enable it.
Room view
kesarAI keeps the last ~30 minutes of a channel (up to about 40 lines) in memory and reads them only when a question needs them — “what do you think of that?”, “who's right?”, “summarise the chat”. Other questions are answered without looking at the chat. Those lines are never stored by kesarAI and never used for AI memory; they go to the AI provider (Anthropic) only when they are read to answer a question. kesarAI's reply may mention what it saw. Like any reply, it stays in the asker's conversation for up to 30 minutes, and it is removed when anyone whose messages it saw forgets their AI data, pauses AI memory or clears their conversation. Room view skips ticket and verify channels, (private) asks and anyone who paused AI memory, and deleted messages drop out immediately.
Room view is used only with Claude. Server administrators can turn room view off for their server (/setup → AI Access → Room View).
Personalities
Pick a personality on your account page:
- Friendly — the same answers, with no jokes.
- Default — short and casual, with the odd light joke, and the right answer first.
- Rude — teases you (only you) while still answering correctly.
- Flavours: Concise, Pirate, Anime, Gen-Z, Shakespearean and Scholarly.
kesarAI only ever teases the person asking — never anyone else, and never about identity, looks or family. Ticket, verify and staff channels are always plain. Type “stop”, “i'm serious” or “not funny” and jokes are off for you in that channel for 30 minutes.
A server admin can turn on Force server personality. On Claude it then sets the most teasing allowed in that server (Friendly < Default < Rude), and its flavour (like Pirate) applies to everyone.
What it can use
Depending on the question, kesarAI can read recent messages in the channel (including who talks the most), look up members (with their level) and roles, search the web, read a link, look at images, transcribe voice messages and YouTube captions, make a poll or chart, speak in your voice channel, play and control music, save something to your memory, or react with an emoji.
- There are no per-user tool switches: kesarAI picks what a question needs.
- While a tool runs, the reply shows a status line such as
Searching web: …, then the answer streams in. - Answers that used tools end with a small line listing them, for example
-# used: web_search, and search answers show up to three sources. - Polls, charts, speaking and music happen only when you ask for them in your own message — text in the channel or in search results can't trigger them. kesarAI also learns facts about you from your public messages to it, privately, unless you pause memory.
- kesarAI won't show messages from people who paused AI memory.
Battery and limits
- Every answer uses a little of your AI battery, at what it really cost to make; searches and
ksr thinkuse more. See AI battery. - An answer that has started always finishes. If the battery is empty, kesarAI tells you when it will be back instead of answering.
- Spam limit: about 10 AI requests per minute per person. Extra requests get a ⏳ reaction instead of a reply. Normal use never hits it.
Privacy
Your message and its context go to the provider of the model you chat with: Google for Gemini (the default), Anthropic for Claude. Search words go to Google on Gemini and to Brave on Claude. Room-view lines and search results are never stored by kesarAI. kesarAI saves facts you state about yourself in public messages to it, and anything you ask it to remember, unless you pause memory. Deleting your AI data doesn't reset your battery. Full details in the Privacy Policy.
The dashboard's Memory page is the web twin of /askmemory on Discord: inspect what kesarAI knows about a user, wipe one user's record, or wipe the whole memory database. Memory is universal across servers, so every action there is global and only the bot owner can take it. More in admin tools.
