Support & FAQ

No. Pick "Continue as guest" and you can start chatting immediately — no email, no signup. An account is optional: signing in with Google or an email code gets you roughly 4x the daily usage, and unlocks signed-in extras like image generation and the Spicy model.

If unsure, start on ROK Hermes Lite and step up only when a reply needs more depth:

• ROK Hermes Lite — the fastest replies, great for everyday questions and quick facts.
• ROK Hermes — the full 120b reasoning model for harder problems, coding, and long answers.
• GLM 5.3 Flash / GPT-5.6 Luna / GPT-5.6 Luna Pro — ROK-managed cloud models; Luna Pro adds extended reasoning.
• ROK Hermes Spicy — the uncensored adult-content mode. Strictly 18+, signed-in users only.

Yes — chatting, the tools, and the developer API are all free. Heavy use is paced by a daily allowance that refills every day at 00:00 UTC, which keeps the shared models fast and available for everyone.

A credit is one upstream model call. A simple reply on ROK Hermes Lite costs 1 credit; heavier models cost more per call (ROK Hermes and the Luna models cost 3). You'll see a small "≈N credits" note under each reply showing exactly what that turn cost.

When the assistant uses tools — searching the web, working with files, or generating images — each step is another model call, so a tool-heavy turn can cost several credits. The per-turn note under the reply shows the total. Quick chats on Hermes Lite stretch the day's allowance the furthest.

New turns stop until the allowance refills at 00:00 UTC — the banner above the composer counts down for you. If you're signed in and using ROK Hermes or Hermes Spicy, ROK automatically continues on the faster Hermes Lite engine instead of stopping you (you'll see an amber notice). Guests hit a hard stop.

Read the Terms of Service

Sign in. Signed-in accounts get roughly 4x the guest allowance, and tool turns cost the same as anyone else's. You can also stretch the free tier by sticking to Hermes Lite for simple questions.

Yes. Image generation and editing draw from their own small daily allowance (4 images a day), separate from your credit budget — generating an image never drains your chat credits. A failed generation refunds the slot automatically.

On your device, in your browser — even when you're signed in. Your chat history is not uploaded to ROK's servers unless you share a conversation or attach files for the assistant to work with.

Read the Privacy Policy

Yes. Open any conversation and use the export action to download it as JSON or Markdown. You can delete individual chats — or wipe everything — at any time; deletion removes the chat from your device's storage.

Sharing creates a public link (with an unguessable ID, not your personal info) that serves a frozen copy of that conversation. Anyone with the link can view it, and you can revoke the link at any time — it stops working immediately.

No. Your chats are not used to train our models, and we don't sell your data. One exception to know about: prompts sent through the uncensored Spicy model are logged together with the account email so prohibited use can be investigated — the Privacy Policy covers the details.

See the Privacy Policy

Beyond plain chat, the assistant can use real tools when a request calls for them:

• Search and read the web for current information.
• Work with files in your workspace: create, edit, search, and present finished files for download.
• Run code and shell commands to build or verify things.
• Generate and edit images (signed-in users).
• Describe uploaded images and read files you attach.

Tool turns run in steps — search, read the results, maybe another step, then answer — and each step is a separate model round. You'll see progress cards as each tool runs, so you can follow what it's doing.

Yes. Settings has a "Custom API key" field — leave it empty to use ROK's hosted models, or plug in your own provider key to run against your account instead of ROK's shared allowance.

This is usually the upstream model provider having a moment, not something you did. Wait a few seconds and retry. If it keeps failing on one model, switch to another (Hermes Lite is the most reliable fallback) — and if it's down everywhere, it'll likely clear shortly.

Short-burst rate limits protect the shared free tiers so no single user can saturate them. They're separate from your daily allowance and clear on their own within minutes — pause briefly and retry.

Chat history lives in the browser you used — it doesn't sync across devices or browsers, and it won't appear in a private/incognito window or after clearing site data. Signing in keys your usage limits, not your history, so an account won't back-fill chats from another device.

Email us and we'll dig in. If the error shows a "ref" code, include it — it maps directly to the failure in our logs — along with your browser and roughly when it happened.

[email protected]

Yes — ROK exposes a free, OpenAI-compatible chat completions API at rokremasteredapi.rokteam.org/v1, with streaming, tool calling, and no API key required to start.

Read the API reference