JanitorAI
By janitor
JanitorAI typical implementation process:
Sign up and create an account (5–10 minutes): Go to JanitorAI’s website, register with email, Google, or Discord, then set up your user persona and basic profile settings.
Configure the AI “brain” (10–30 minutes): Open any character chat, go to API Settings, and choose your model—either free JanitorLLM (no key needed) or an external LLM like OpenAI/DeepSeek via API key or proxy configuration.
Create or select characters (15–45 minutes initially): Browse trending/public characters or build your own by defining name, bio, personality, scenario context, and initial message, then save and start a chat to test behavior.
Refine prompts, memory, and settings (ongoing over a few sessions): Iterate on character descriptions, safety/NSFW toggles, and system prompts to improve consistency and tone, typically across a few hours of experimentation spread over days.
JanitorAI is highly customizable and can be adapted to many specific business workflows rather than just casual role‑play.
Custom business personas and scripts: You can build characters that behave like customer‑support agents, sales reps, HR assistants, trainers, or internal ops bots by defining their goals, tone, escalation rules, and example dialogs.
Rule prompts and boundaries: Detailed instruction blocks let you enforce brand voice, compliance constraints, do/don’t lists, and escalation behavior, which is critical for customer service, lead qualification, or internal knowledge assistants.
Use‑case specific scenarios: Characters can be placed in defined scenarios (e.g., “first‑line SaaS support,” “B2B SDR doing discovery,” “HR onboarding coach”) so their responses stay aligned with your business context and workflows.
API integration into systems: Through its API pattern, JanitorAI can be wired to external services (CRMs, ticketing, internal tools) so bots can receive structured inputs and return formatted outputs for things like triaging tickets, data cleaning, or routing requests.
Data and workflow automation: Guides show JanitorAI used for web‑scraping cleanup, formatting CSVs, summarizing records, drafting emails, and automating repetitive ops tasks, effectively acting as a no‑code assistant on top of your data sources.
Training and role‑play simulations: Businesses use customized characters to simulate customers, interviewers, or managers for soft‑skills training, objection handling practice, and language or scenario training.
Visual and profile customization: CSS profile customization allows branded character hubs or internal “bot catalogs” that match company aesthetics, making it easier to onboard teams to the right assistants.
JanitorAI offers mainly self‑serve training resources, an in‑app help bot, and a very active community rather than formal enterprise onboarding.
For new users, the platform maintains a dedicated help center with user guides, creator guides, FAQ, and troubleshooting sections, covering topics like using JanitorLLM, setting up proxies, writing custom prompts, and building bots step by step. There is also a curated “Welcome to the JanitorAI Help Desk” hub that explains how to use the knowledge base, links to more than 70 articles, and introduces a customer‑service chatbot trained to answer policy questions, give CSS and character‑building tips, explain LLMs and proxies, and assist with common issues.
Beyond official documentation, JanitorAI is heavily supported by community content and live discussion channels. Power users maintain long‑running master guides, bot‑creation tutorials, and advanced prompting/proxy walkthroughs on Reddit and external blogs, while the official (age‑gated) Discord server provides real‑time help from other users, moderators, and sometimes team members, with channels for troubleshooting, API configurations, character design, and announcements.
JanitorAI uses some standard security practices to protect data, but most independent reviews stress that it is not a high‑security, enterprise‑grade platform and should only be used for non‑sensitive information.
JanitorAI’s own privacy policy and external analyses indicate that it applies “appropriate technical and organizational measures” to protect personal data against unauthorized processing, accidental loss, destruction, or damage, including encryption for data in transit and at rest. Some sources describe the use of end‑to‑end encryption for chats and AES‑256 encryption for stored data, relying on major cloud providers such as AWS or Google Cloud for physical and infrastructure security. Chats are private by default and are not publicly visible unless a user explicitly enables options like “Make chat public,” and the service states it does not automatically collect real‑world identity data unless users share it in conversation or registration fields.
At the same time, security commentators highlight important limitations and risks. Public documentation around long‑term data retention, detailed encryption architecture, third‑party access and sharing with “AI model partners,” and deletion tooling is much thinner than on enterprise AI platforms, and there have been reports of issues such as expired SSL certificates on login endpoints and limited GDPR‑style controls. Analyses emphasize that conversations and metadata (IP address, device info, usage patterns) are logged for service optimization and that data may pass through multiple systems, including external LLM APIs or reverse proxies if users configure those, which increases the attack surface. Several security guides therefore conclude that JanitorAI is generally safe only for casual, creative use and roleplay, and explicitly advise against sharing personal, financial, or confidential business information or using it for regulated or highly sensitive workloads.
JanitorAI’s privacy documentation and independent reviews show that the platform retains broad control over how chat data is stored and used, and offers only limited, mostly manual options for data portability.
In terms of ownership, the official privacy policy frames JanitorAI itself as the “Controller” of personal data, which means the service—not the user—formally decides how data is processed within the platform, even though users retain practical control over what they choose to submit. The policy states that JanitorAI collects email addresses, usage data, and user‑generated content (including characters and chat logs), and may store chats temporarily to improve AI performance, support character memory, and provide an immersive experience. External analyses highlight that there is no detailed, public statement that explicitly says “you own your content and can freely export everything,” and that retention periods for chat logs are not clearly specified.
On portability, users can copy or manually save individual conversations and character definitions, but there is no simple “export all my data” function documented for regular users, and community threads confirm that bulk export of all chats is not currently supported. Guides therefore recommend treating JanitorAI as a transient, experience‑centric tool: if you want durable ownership of characters or key outputs, you should regularly back them up yourself, for example, by exporting personas as JSON or character‑card formats into other tools rather than relying on JanitorAI as your long‑term content repository.
JanitorAI operates more like a consumer web app than an enterprise SaaS, so renewal and cancellation terms are relatively simple but not very formally documented.
Auto‑renewal and billing cycle: Third‑party guides and reviews describe JanitorAI Pro/paid plans as standard recurring subscriptions (monthly or yearly) that auto‑renew by default until you downgrade or cancel from your account settings. Users are advised to track their next billing date and turn off renewal in advance if they do not want to be charged for the next term, since automatic renewal is enabled out of the box.
Cancellation process: Tutorials explain that you cancel by going into Account or Billing settings, locating your subscription, and using a “Cancel account/plan/terminate subscription” style button, then confirming cancellation on the web interface. After cancellation, you typically keep paid features until the end of the current billing period, at which point the account falls back to the free tier while your login and basic data remain unless you explicitly delete the account.
Refund windows (informal guidance): Detailed third‑party breakdowns of JanitorAI’s paid plans describe a pragmatic but not contractually guaranteed refund approach: monthly subscribers can cancel anytime, with a small grace window (for example, roughly the first 72 hours and only if they have not used a large portion of their message quota) where support may approve a refund; yearly subscribers are said to have a short “cool‑off” window (around 14 days) in which a prorated refund may be granted if the platform clearly is not a fit. Outside those windows, cancelling normally just stops the next renewal and leaves you on the plan until expiry, with no refund for unused time.
