How to Use Chutes on Janitor AI in 2026

Chutes are saved message sequences that allow you to fire off a pre-written response in a roleplay conversation with a single click. Think of them as quick-draw replies for specific scenarios—like a greeting, a combat action, or a flirtatious advance. Each chute contains one or more lines of text that the Janitor AI interface sends as your character's message. They are stored in your account and can be loaded on any compatible character chat. Chutes work with both the free Janitor LLM and external APIs like OpenAI or Kobold. They are particularly useful for repetitive opening lines or establishing consistent character behavior across multiple roleplay sessions. Unlike persona settings which affect the AI's behavior, chutes control your own character's dialogue. They are essentially macros for roleplay, reducing typing time and ensuring consistency in how you portray your character.

Loading a chute is straightforward once you're inside a character chat. First, open the chat interface for any Janitor AI character. Look for the 'Load Chute' button, typically located near the message input box or in the toolbar above it. Clicking this button opens a dropdown or popup listing all your saved chutes. Select the chute you want to use—it will automatically insert the chute's content into the message input field. You can then edit the text before sending if desired, or send it as-is by pressing Enter. If you have no chutes saved, the list will be empty. Chutes are specific to your account, not to individual characters, so any chute you create can be used in any chat. This feature is available on the web version and mobile app. Note that chutes do not affect the AI's response generation—they only pre-fill your message.

To create a chute, navigate to your Janitor AI profile or settings page. Look for the 'Chutes' or 'My Chutes' section—usually found under account settings. Click 'Create Chute' or a similar button. You'll be prompted to enter a chute name (e.g., 'Greeting' or 'Attack'), and then type the message content in a text box. You can write multiple lines or even paragraphs. There is no character limit mentioned on the platform, but practical use suggests keeping chutes under 500 characters for compatibility. After saving, the chute appears in your list. You can also edit or delete chutes from the same section. A practical tip: create a set of chutes for common roleplay situations like introductions, negotiations, or combat moves. This saves time and keeps your character's voice consistent. You can have as many chutes as you need—there's no published limit.

Chutes function identically whether you use Janitor AI's built-in LLM or connect an external API. When you load a chute, it inserts text into the message box regardless of the backend. However, the AI's response quality depends on the LLM you've selected. For users who connect OpenAI or Kobold via an API key, chutes are especially useful because they let you quickly send complex prompts that steer the AI's behavior. For example, a chute containing 'The knight draws his sword and says, "You shall not pass!"' can set the scene without typing. The external LLM will then respond based on that input. Note that chutes do not bypass any API usage costs—each message you send (including chute content) counts toward your API token consumption. For heavy chute users, it's wise to monitor your API usage. Janitor AI does not restrict chute functionality based on the LLM choice.

Users sometimes report that chutes fail to load or appear blank. The most common cause is a browser cache issue—try refreshing the page or clearing cache. Another issue: if you created a chute on one device, it may not sync immediately to another; log out and back in to force sync. Some users find that chutes with special characters (like asterisks or quotes) break the insertion—avoid using unescaped HTML or punctuation. Also, chutes are not automatically saved across sessions if you didn't click 'Save' after editing. If a chute loads but doesn't send, check your internet connection; Janitor AI requires an active connection. For API users, ensure your API key is still valid and has sufficient credits. Finally, remember that chutes are client-side tools—they don't interact with the AI's memory or context beyond the message you send. If the AI ignores your chute's content, it's likely due to the model's context window or prompt formatting.

If chutes feel limited or you want more advanced automation, consider using persona settings or custom system prompts in Janitor AI. Personas allow you to define your character's traits that the AI remembers across messages. For more complex automation, some users write scripts using browser extensions like Tampermonkey to auto-send messages. However, these are not officially supported. Another alternative is to use a different platform altogether. For example, AIAngels offers a built-in custom companion builder with no need for external APIs or chutes—your companion remembers everything from conversation 1 without any setup. AIAngels also includes unlimited text chat on its free tier, no message caps, and no credit systems. While Janitor AI's chutes are a workaround for consistency, platforms like AIAngels bake persistent memory and customization directly into the experience, removing the need for such macros.