What Janitor AI and HoneyChat offer
What is there
User bots
Cards with a personality, a scenario and a first message; authors get analytics (since 13.01.2026), verification and scheduled releases with a countdown (since 19.01.2026).
Public chats
Since 31 March 2026 any chat can be published: others read, react, fork and continue your story.
Multiple personas
Switch between your personas right in the dialog — the bot treats them as different people.
Your own API
API settings with several saved prompts and answer-length control (redesigned 19.07.2026) — plug a third-party model in with your key.
janitor+ Router
Since 30 July 2026 — access to third-party models through Janitor itself, with transparent prices and $5 of starting credits for subscribers.
Short replies
A short-answer model opened to everyone on 15 July 2026.
Models: three ways
| Path | What it is | Price | Source |
|---|---|---|---|
| JanitorLLM | The service's own free model; v2 preview since 10.10.2025 | $0 | Newsroom, 10.10.2025 |
| janitor+ | 5× context, priority and a “generous monthly pack of swipes on frontier models” | $12.99 a month | Announcement of 24.06.2026 |
| janitor+ Router | Third-party models through Janitor “with transparent prices”, $5 of starting credits for subscribers | at the models' rates | Announcement of 30.07.2026 |
| Your own API | A third-party provider's key in the API settings; several saved prompts | at the provider's rates | Newsroom, 19.07.2026 |
Which exact “frontier” models are included in the janitor+ swipes and which are available through Router the service does not list — not verified.
The free JanitorLLM is “what there is”; the community regularly complains about repetitiveness, and v2 was announced precisely as an answer to that. A good experience in Janitor has traditionally been built on your own key to a third-party model — a separate bill, a separate provider account and prompt tuning. Router simplifies the path but stays paid on top of the subscription.
Memory
In Janitor “memory” is the model's context window, and it depends on which model you use. On JanitorLLM it is the base size (the number is not published — not verified), with janitor+ — “5× more context for improved memory during chats”. With your own API key the context is set by the provider. A separate long-term layer of facts “about you” that survives the context is not described in the announcements.
- Long bot and persona cards take part of the context before the dialog even starts.
- On the free model the window is smaller; at peak hours a queue is added.
- Keep key facts in the persona description and the bot card — they are fed to the model every time.
What is missing
- Voice. No voice notes, no calls.
- Images and video. No generation in chat; in the janitor+ announcement these are “plans” — “think voice and image”.
- Access without an account. Chats require registration.
- A ready “companion”. Memory, model and style are configured by the user, not set by the product.
The case for Janitor AI
More tools for role-players and authors than anyone in the niche: public chats, forks, personas, release scheduling, custom CSS, your own API. It is a platform, not one character.
The case for HoneyChat
Everything Janitor lacks is included in the plan here: expressive voice notes, photos in the context of the scene, short videos, memory with facts about you and a story journal. And nothing needs configuring — the character is ready at once, in Telegram.
