Janitor AI pricing

About

Janitor AI is a consumer AI character chat and roleplay platform — users talk to user-created AI characters, write collaborative fiction, and run open-ended roleplay sessions. It launched in June 2023 and went viral almost immediately, reaching roughly 1 million users in its first 7 days. By 2026, coverage references a user base in the tens of millions, skewed heavily toward Gen-Z, with an unusual demographic: roughly three of every four users are women.

The company is privately held and was founded by Jan Zoltkowski. Its defining moment came in July 2023, when OpenAI issued a cease-and-desist and cut off API access over content-policy concerns. Rather than fold, Janitor built its own hosted model — JanitorLLM (JLLM) — and added a “proxy” system letting users plug in their own external API keys. That fork in the road is the whole reason its pricing looks the way it does today.

For the most current information, visit Janitor AI.

Pricing summary : How Janitor AI’s pricing model works

Here’s the thing most write-ups get wrong about Janitor AI: the app is free, but the model behind it is where the money is — and most of that money never touches Janitor’s bank account.

There are three ways to power your conversations:

1. Free JanitorLLM (JLLM) — Janitor’s own hosted model, in beta. No API key, no cost, capped at roughly 50 messages/day. This is what the vast majority of users run on.
2. Pro — $9.99/month or $99.99/year (as of April 2026) — the first-party paid tier. It lifts message limits, adds priority queueing and peak-hour stability, and unlocks access to all models. This is Janitor’s only direct revenue line from end users.
3. Bring-your-own external API key — connect OpenAI, Anthropic Claude, or OpenRouter via a “proxy” in API Settings. You pay that provider per token. Janitor captures $0 of this — it’s a pure pass-through cost on someone else’s bill.

So the question “how much does Janitor AI cost?” has three honest answers: $0, $9.99/mo, or “whatever your OpenAI/OpenRouter usage runs you” (typically $10–$30/mo for active users).

What makes this different: Janitor decoupled the experience from the inference. Unlike a typical AI product that marks up tokens and resells them inside a subscription, Janitor lets the user own the relationship with the model provider. The platform monetizes a thin slice (Pro) while the dominant cost driver — frontier-model tokens — is externalized entirely to the user.

Pricing by product

Sales motions across products: pure self-serve PLG. There is no sales team, no enterprise quote, no contract — you sign up and start chatting. Pro is a one-click consumer subscription; the external-key path is a DIY config screen.

External-provider reference points (these are the third party’s prices, not Janitor’s): OpenRouter gives 50 free messages/day on free models, and a one-time $10 credit top-up permanently raises that to 1,000 messages/day. OpenAI new accounts get a small trial credit (around five US dollars, ~500 messages) before per-token billing kicks in. Anthropic Claude is pay-as-you-go per token.

Hidden costs : What Janitor AI users actually pay

The “free” framing hides the real bill for serious users. The free JLLM tier is genuinely free — but it’s beta-quality, rate-limited, and prone to downtime and “No response from bot” errors during peak load. The moment you want reliability or a smarter model, you either pay Janitor $9.99/mo or you start paying a third party per token. That second path is the hidden cost, because it scales with how much you chat and shows up on a bill Janitor never itemizes for you.

The traps to watch: token costs on a frontier model scale with conversation length — long roleplay sessions burn context every turn, so a chatty week can spike your OpenAI/Claude bill well past the headline estimates. And the Pro subscription does not eliminate external-key costs if you specifically want GPT-4- or Claude-grade output — Pro improves access to Janitor’s hosted models, not your third-party token spend.

Want to estimate your own Janitor AI bill? Use the Janitor AI pricing calculator to model your costs based on usage patterns.

Pricing evolution : Janitor AI pricing history and changes

Cadence

Tracked range: 2023–present.

Notable changes

• June 2023 — Launches free, routing chats through OpenAI’s API. Explodes to ~1M users in a week.
• July 2023 — OpenAI cease-and-desist cuts API access. Janitor builds proprietary JanitorLLM and ships bring-your-own-API-key proxy support. The “free app, you pay for the model” structure is born here.
• 2024–2025 — Long stretch with essentially no first-party paid tier; revenue model leans on the externalized-token approach and JLLM remains free-but-beta.
• April 2026 — Pro at $9.99/mo ($99.99/yr) confirmed in market — the first durable first-party subscription, monetizing priority access to JLLM.
• 2026 — JLLM V2 announced on new NVIDIA B200 GPU clusters (wider context, better formatting), responding to community complaints that JLLM had grown “repetitive.”

What’s unique : Janitor AI’s distinctive pricing mechanics

1. The biggest cost line is externalized to the user. Most AI products bundle inference into a subscription and mark it up. Janitor inverts this: the heavy frontier-model spend sits on the user’s own OpenAI/Claude/OpenRouter account. Janitor’s margin isn’t tokens — it’s the thin Pro layer plus its own cheaper-to-run JLLM.

2. Bring-your-own-model as a first-class pricing path. The “proxy” config makes the user the buyer-of-record for inference. This is closer to how a developer tool works than a consumer app — except the audience is mainstream Gen-Z roleplayers, not engineers.

3. A free hosted model as the default, not a teaser. JLLM isn’t a trial that nags you to upgrade — it’s the path most users stay on indefinitely. Pro and the BYO-key route are opt-in quality/reliability upgrades, not gates on core usage.

Strengths & weaknesses

Billing UX : Janitor AI billing controls and transparency

• Billing controls — Pro is a standard consumer subscription with a 72-hour refund window (if the balance hasn’t been used extensively). The external-key path has no Janitor-side billing at all — spend controls live in your OpenAI/OpenRouter/Anthropic dashboard, not in Janitor.
• Usage visibility — Free-tier message caps (~50/day) are visible in-app, but token spend on an external provider is not surfaced inside Janitor — you have to watch the provider’s own usage dashboard. This is the main predictability gap.
• Payment options — Pro is paid directly to Janitor; external-model usage is paid to the third-party provider. The official pricing page is effectively bot-walled, so most users discover exact figures through community guides rather than a first-party rate card.

Strategic wins : Why Janitor AI’s pricing decisions worked

1. Survived a supplier cutoff by externalizing the model

When OpenAI revoked access in 2023, a token-reseller business would have died. Janitor’s response — build JLLM and let users bring their own keys — turned a single-point-of-failure dependency into optionality. See how AI companies structure pricing.

2. Kept COGS off its own books

By pushing frontier-model spend onto users’ accounts, Janitor avoided the margin trap that crushes many consumer AI apps reselling marked-up tokens. Its costs are dominated by its own cheaper JLLM infrastructure, which it controls. Related: bill shock and cost unpredictability and outcome-based pricing trends.

3. Free-first growth, monetize reliability

A free default model drove enormous scale; the $9.99 Pro tier and BYO-key path monetize the minority who need more without taxing the majority. See choosing the right usage metric.

Areas to improve : Gaps in Janitor AI’s pricing approach

1. Token spend is invisible and uncapped

Because external-model usage is paid off-platform, users get no in-app meter, budget, or alert. Long sessions can quietly run up a provider bill. A native spend dashboard or cap would close a real bill-shock gap. See bill shock and cost unpredictability.

2. The rate card isn’t first-party transparent

The official pricing page is hard to reach for non-logged-in users and bot-walled to crawlers, so exact Pro pricing circulates through third-party reviews. A clean, public rate card would reduce confusion and improve trust.

3. Pro doesn’t fully solve quality

Paying $9.99 improves access to JLLM, but users chasing GPT-4/Claude-grade output still need a separate paid key. A bundled premium-model option (Janitor-resold, predictably priced) could capture spend currently flowing entirely to OpenAI/OpenRouter.

Key takeaways

1. The app is free; the model is the cost. Janitor’s whole pricing story is the decoupling of the chat experience from the inference behind it.
2. A first-party paid tier now exists. Pro at $9.99/mo / $99.99/yr is real as of 2026 — earlier “Janitor has no subscription” framing is now outdated.
3. The dominant cost line lands on someone else’s bill. Power users pay OpenAI/Claude/OpenRouter directly ($10–$30/mo typical) — a pass-through Janitor never captures.
4. Free isn’t free of friction. JLLM’s beta reliability is the reason users pay at all — Janitor monetizes reliability, not access.
5. Externalizing tokens is a durable model for a consumer AI app — it sidesteps the negative-margin token-reseller trap and survived a supplier cutoff.

UBP implications

1. You can sell experience while letting the customer own the meter. Janitor proves a consumer audience will configure their own API key and absorb per-token billing if the core product is good and free. For the trade-offs of consumption versus subscription framing, see introduction to usage-based pricing.
2. A free hosted model is a moat and a margin lever — owning cheaper inference (JLLM) lets you keep a free default that a token-reseller couldn’t afford.
3. If the heavy usage cost is off-platform, surface it anyway. The biggest UBP lesson here is the missing one: invisible third-party token spend is a trust gap any platform externalizing usage should close with in-app metering.

Sources

• Janitor AI homepage (accessed 2026-06-16)
• Janitor AI pricing page (bot-walled; accessed 2026-06-16)
• Janitor AI proxy / API-key setup guide (accessed 2026-06-16)
• Third-party pricing reviews: ToolJunction, Scribe, Proxidize (accessed 2026-06-16)

Bottom line

Janitor AI’s pricing is a study in decoupling: the chat is free, and the real cost — frontier-model tokens — is pushed onto the user’s own OpenAI/Claude/OpenRouter account. A first-party Pro tier ($9.99/mo or $99.99/yr) now monetizes reliability and priority access to Janitor’s own JanitorLLM, but the platform deliberately captures little of the heavy inference spend. It’s one of the few consumer AI products that survived losing its model supplier by making the user the buyer-of-record for inference. Browse the pricing blueprint for more fully-researched company profiles.

Want to compare Janitor AI against other consumer AI platforms? Browse the pricing blueprint.

Pricing timeline : Major events on a vertical axis

Each milestone below corresponds to a public pricing change, product launch, or material adjustment. Major events use a filled marker; minor adjustments use a faded one.

Pro subscription at $9.99/mo

Apr 2026

First-party paid tier confirmed: Pro at $9.99/month or $99.99/year unlocks higher message limits, priority queueing, peak-hour stability, and all models. Free JLLM tier remains.

OpenAI cease-and-desist → proprietary JLLM

Jul 2023

OpenAI cuts off API access over content policy. Janitor builds its own hosted model, JanitorLLM (JLLM), and adds bring-your-own-API-key proxy support so users can plug in OpenAI/Claude/OpenRouter themselves.

Launch — free app on OpenAI's API

Jun 2023

Janitor AI launches as a free AI character chat platform, initially routing conversations through OpenAI's API. Hits ~1M users in 7 days.