Built-in AI: use ReplyRabbit without your own API key
Built-in AI is an AI provider option that ReplyRabbit runs for you, so a Pro or Team licence can draft replies without signing up for an AI provider or pasting in an API key. It comes with a fair-use monthly allowance shared across your licence. Once it's used up, built-in AI pauses until the allowance resets, and you can switch any mailbox to your own key at any time.
Built-in AI is being switched on gradually
Which plans include built-in AI
Built-in AI is part of the paid plans. The Free tier uses your own OpenAI key only.
| Plan | Built-in AI | Draft replies a month |
|---|---|---|
| Free | Not included | Not applicable |
| Pro | Included, on the site your licence is activated on | Up to 150 |
| Team | Included, shared by every site and mailbox on the licence | Up to 500 |
The draft figure is a ceiling, not a promise. Built-in AI also has an overall monthly allowance that covers every built-in AI feature, such as drafts, next actions, AI Signals, attachment understanding, and embeddings for search. Long conversations and many AI features use it faster, so you can reach the allowance before you reach the draft limit.
How the monthly allowance works
The allowance belongs to your licence, not to one mailbox. Every mailbox set to built-in AI draws from the same pool, and on Team that includes every installation on the licence.
The allowance resets once a month on your licence's billing cycle, so the reset date isn't always the first of the month.
To check your usage, open a mailbox's ReplyRabbit settings with built-in AI selected. Under AI Provider, ReplyRabbit shows:
- Drafts this month: how many built-in AI drafts you've used out of your plan's limit.
- Shared allowance remaining: the percentage of the overall allowance that's left.
- Shared allowance status: shown instead of a percentage when ReplyRabbit can't read the remaining amount.
- Resets: the date the allowance starts again.
Local token analytics and any Monthly Token Budget you set on a mailbox are separate from this licence-wide allowance.
What happens when the allowance runs out
ReplyRabbit stops sending built-in AI requests until the allowance resets. It doesn't switch to another provider on its own, and it doesn't charge you extra.
You'll see one of two messages on the draft and in mailbox settings:
- Drafts used up: "Built-in AI drafts are used up until" a date, followed by a suggestion to switch to your own AI provider in Mailbox Settings to keep drafting.
- Overall allowance used up: "The built-in AI allowance is used up until" a date, followed by a suggestion to switch to your own AI provider to keep using AI features.
If the reset date isn't available, the message says "until it resets" instead. Once the allowance resets, built-in AI picks up again with no action from you.
Switch a mailbox to your own API key
The AI provider is set per mailbox, so you can keep built-in AI on one inbox and use your own key on another.
Open the mailbox's ReplyRabbit settings
In FreeScout, open the mailbox, go to its Settings, and click ReplyRabbit.Choose your provider
In the AI Provider card, under Drafts, change Provider from ReplyRabbit managed AI to the provider you want, such as OpenAI, Anthropic, or Ollama.Add your key and pick a model
Paste your API key (local providers don't need one) and choose a model. Built-in AI picks its own models, so there's no model to carry over.Save
Save the settings. New drafts for this mailbox now use your provider, and your provider bills you directly for that usage.
A few things to know when switching either way:
- Per-task providers stay put. If you've chosen a specific provider for a task like previews or attachment understanding, that task keeps using it. See Per-task AI models.
- Embeddings are provider-specific. Search embeddings made by one provider can't be reused by another, so changing the embedding provider means your sources are embedded again, which uses the new provider's embedding credits. If the indexed counts look low afterwards, use Refresh Index on that source.
- Conversation semantic search uses keywords on built-in AI. While built-in AI is selected, conversation semantic search falls back to keyword search, because its separate provider can't use the built-in allowance.
To compare providers before you switch, see Choosing an AI provider.
Privacy: where built-in AI sends support content
With built-in AI, your FreeScout server sends the content needed for the AI features you've turned on, such as customer messages, your replies, image attachments, and text being indexed for search, to the ReplyRabbit AI gateway. Moogento runs the gateway on Cloudflare. The gateway forwards each request to OpenRouter or directly to an AI model provider.
What that means for your content:
- No training: the AI model providers don't train on your prompts or outputs.
- Limited retention: a provider may keep prompts and outputs for up to 30 days, or longer where its safety systems flag them or the law requires it.
- Providers may vary: the model and provider used can change, and if a provider can't take a request, it can move to another one.
The gateway keeps AI results encrypted for up to 30 days, so a retried request isn't processed and counted twice. To open a session, ReplyRabbit also sends your licence key and site URL to the gateway, where they're stored encrypted.
If support content must never leave your own server, use a local provider such as Ollama instead. For the full details, read the Privacy Policy.