Updated: 2026-07-24
A roleplay chat bot is a conversation where the AI does not "help you step by step" but plays a character: it holds a personality, a way of speaking and a first-person role. To start an rp chat with AI you have two paths — pick a ready character from the catalog, or build your own by setting a personality, speech rules and a greeting. How convincing the roleplay is comes down not to "model power" but to the system prompt: it is what tells the bot who it thinks it is.
Pick a character or start your own. Open the "Characters" section and take a ready preset — or start from scratch via agent creation if you want a character of your own.
Set the personality. Fill in the system prompt: who the character is, from which world, what matters to it, which topics it avoids. Or describe the role in words — the setup AI assistant drafts the prompt for you to adjust.
Write the speech rules. In the rules field set the manner: reply length, formal or casual address, signature phrases, a ban on breaking character.
Write the greeting. The first message sets the scene and tone — it tells the other side right away who they are with and where they landed.
Test the roleplay in the preview. Send the character a few lines in the preview block. Throw an awkward question at it — make sure it holds the role instead of slipping into an assistant tone.
Open the roleplay where you need it. Turn on the public page for roleplay by link, connect a Telegram bot with a token from @BotFather, or put the body on the desktop — one character, many channels.
An ordinary chat bot answers "like an assistant": neutral, in bullet points, from a "how else can I help" stance. In a roleplay chat the bot stays in character — it speaks as the character, keeps its personality and speech style across the conversation, and does not drop into a service tone at the first awkward question.
In Corpis a character also has a visible body — a vector figure that reacts with emotion to how the conversation goes (joy, thought, worry). It is not an avatar: the reaction helps the roleplay land instead of staying as bare text in a window.
The fast start is the "Characters" section: ready presets from anime, game and film universes. Each character already has a written personality, speech rules and a first message, and the body is picked from franchise looks. You open a character, see how it speaks, and create such an agent in a minute.
The preset is not all-or-nothing: the character's system prompt is shown openly, and during creation you can rewrite it into your own version of the roleplay — change the tone, add facts about the world, adjust the speech rules.
Your own character is created with a form, no code: name, personality (system prompt), behavior rules and a greeting. You do not have to write the prompt by hand — describe in words who the character is and how it speaks, and the setup AI assistant fills the fields for you to adjust.
A separate branch is the clone from chat: if you need the roleplay of a real speech style, a Telegram chat export is parsed right in the browser and turns into a character speaking in that voice. That is a different wizard with its own pipeline — here we mean playing a fictional or made-up role.
The system prompt is the personality: the agent's brain receives it before every reply, describing who the character thinks it is, what it knows and how it speaks. A vague prompt ("be a friendly elf") gives vague roleplay; a specific one (who the character is, where from, what matters to it, which topics it avoids, how it builds phrases) holds the role steadily.
Speech rules sit in a separate field and act as a style anchor: short or long replies, formal or casual address, signature phrases, refusing to break character. The greeting sets the tone from the first message — it tells the other side right away who they are talking to.
The brain, meanwhile, is swappable: the model is chosen in the agent settings and changes without rewriting the character — personality, rules and body are stored apart from it.
The same character works across every channel: in the on-platform chat, on the agent's public page (where the other side writes without signing up), in a Telegram bot, in a site widget and in the Windows desktop overlay. Roleplay rules are set once and apply everywhere.
The platform tone is neutral: this is character roleplay and role scenes, not a promise of a "chat without limits". Anything reaching public surfaces passes an automatic content check before publishing — it judges dangerous content, not the fact of roleplay itself. For private roleplay with your own agent this limit does not get in the way.
Just pick a ready character in the "Characters" section and start talking, or build your own with a form: name, personality, speech rules and a greeting. No code is needed, and there is a free tier plus a 3-day Pro trial with no card.
Mostly the system prompt — it is the character's personality, which the brain receives before every reply. A specific description (who the character is, what it knows, how it speaks, which topics it avoids) holds the role steadily, and speech rules lock in the manner. The model is swappable and does not rewrite the character.
Yes. Public characters in the catalog have a page where you can message them anonymously, with no account (a demo limit applies). An account is only needed when you want to build your own character and connect it to your channels.
No. Corpis is about character roleplay and role scenes in a neutral tone, not a promise of a chat without limits. Public surfaces pass an automatic content check before publishing: it filters dangerous content, not the fact of roleplay itself.