ReplyRabbit FAQ
Common questions about ReplyRabbit, the AI draft-reply module for FreeScout. Short answers first, with the details below.
Does ReplyRabbit send my customer data to third parties?
Only for the AI features you enable, such as drafting replies, suggesting next actions, or classifying AI Signals. With your own key, content goes only to the AI provider you choose, and with a local provider like Ollama or LM Studio, nothing leaves your own server. With optional built-in AI on a paid plan, content goes through the ReplyRabbit AI gateway on Cloudflare to OpenRouter or directly to an AI model provider. Providers may keep prompts and outputs for up to 30 days (longer where their safety systems flag them or the law requires it) and don't train on them. Encrypted AI results are kept for up to 30 days so retries aren't processed twice. See the Privacy Policy for details.
Is there a free version?
Yes. The Free tier uses OpenAI, supports one mailbox, and includes 10 AI drafts per month. Paid tiers add more providers and remove the draft limit, and Team adds unlimited mailboxes.
Which AI providers are supported?
OpenAI, Anthropic, Google AI Studio, OpenRouter, Z.AI, Ollama, and LM Studio. The last two run locally on your own hardware.
Do I have to write replies myself?
No. ReplyRabbit drafts a reply that you review and send. You stay in control: nothing is sent automatically.
Who pays for the AI usage and tokens?
With your own provider API key, you pay that provider directly for usage, and local providers like Ollama and LM Studio have no per-token cost. Paid plans can instead use built-in AI, which includes a fair-use monthly allowance shared across every site on your licence. That's up to 150 draft replies a month on Pro and up to 500 on Team. Other AI features, such as next actions, signals, summaries and search indexing, stay within fair use of the same allowance, and the allowance resets monthly. When it's used up, built-in AI pauses until the allowance resets, or you can switch to your own key. Own-key mode works exactly as before.
Does ReplyRabbit use prompt caching?
Yes, where the selected provider supports it safely. ReplyRabbit handles prompt caching automatically, and Monthly Token Budget still counts total input and output tokens even when a provider bills cached input differently.
Can ReplyRabbit run fully offline and privately?
Yes, on a paid tier with a local provider such as Ollama or LM Studio running on your own server. With a local provider, no conversation content leaves your hardware.
How much of my past conversation history does ReplyRabbit use?
On Pro and Team, each draft searches the last 120 replies your team sent to that customer, across every mailbox and with no age limit, and uses the 4 most relevant passages. That limit is per customer, not for the helpdesk as a whole. Drafts also follow up to 8 style examples learned from your edits. Replies sent to other customers aren't used unless you switch on Learn from past replies (Beta, Pro and Team), so general answers still belong in the Knowledge Base, a documentation source or Company Context. See How ReplyRabbit learns from your past conversations in the docs.
What are "next actions"?
A suggested checklist of follow-up steps for a conversation, so you can see what to do next at a glance.
Can ReplyRabbit work with FreeScout Workflows?
Yes. ReplyRabbit can apply Workflow Tags such as priority, intent, no-reply, draft-ready, and Linear issue tags. FreeScout Workflows recipes can use those tags when your installed Workflows module reacts to tag changes.
Does it work in languages other than English?
Yes. ReplyRabbit can draft in the customer's language or in the language you set in your mailbox settings.
How do I get help?
Email us at hey@replyrabbit.app and we will be happy to help.