OpenAI integration
Use this guide to connect OpenAI to ReplyRabbit for a FreeScout mailbox. OpenAI draft generation is available on the ReplyRabbit Free tier. OpenAI can also power next actions, AI Signals, semantic retrieval, and other AI-assisted support text when those features are licensed and enabled.
Looking for the overview?
Overview
OpenAI is ReplyRabbit's Free tier AI provider and remains available on paid plans. You choose it in the mailbox AI Provider settings, paste your OpenAI API key, choose a model option, and save the mailbox.
ReplyRabbit stays responsible for the FreeScout workflow: company context, response style, examples, token budget checks, and reviewable draft output. OpenAI usage is billed separately by OpenAI through the provider account connected to your API key.
For direct OpenAI generation requests, ReplyRabbit sends a safe prompt cache key when mailbox context is available. This gives the provider a stable routing hint for repeated support prompts without exposing your API key. Monthly Token Budget still counts total input and output tokens conservatively.
Prerequisites
Before you start, make sure you have:
- A working FreeScout installation.
- ReplyRabbit installed and activated.
- An OpenAI account with API access.
- An OpenAI API key for the account you want ReplyRabbit to use.
OpenAI draft generation is available on the ReplyRabbit Free tier. Semantic retrieval, additional providers, and other advanced features require Pro or Team.
If you are still choosing a provider, read Choosing an AI provider.
Permissions and API keys
For draft generation, ReplyRabbit uses your OpenAI API key when the mailbox provider is set to OpenAI. The key is stored as a secret and shown masked after saving.
If semantic retrieval is licensed and enabled, OpenAI can also be selected as the embedding provider. In that case, ReplyRabbit may use a dedicated embedding API key, or it may fall back to the OpenAI draft key when the draft provider is OpenAI.
If the key is revoked, rotated, rate-limited, or out of provider credit, update the mailbox settings before generating more drafts.
Connect OpenAI
Open ReplyRabbit settings
In FreeScout, open the mailbox, go to Settings, and choose ReplyRabbit.Choose OpenAI
In the AI Provider card, select OpenAI as the provider.Paste the API key
Paste your OpenAI API key into API Key.Choose a model
Use the latest option unless your team has chosen a specific OpenAI preset or model ID. Use Custom only when you know the exact provider-supported model ID.Set a token budget if needed
Use Monthly Token Budget if you want ReplyRabbit to stop generating after a configured monthly token limit for the mailbox.Save and test
Save settings, open a conversation, and generate a draft. Review the answer before sending it to a customer.
Configuration notes
- Provider: choose OpenAI in the AI Provider card.
- API key: use a key from your own OpenAI account for draft generation.
- Model: choose latest, a listed preset, or a custom provider-supported model ID.
- Prompt cache routing: ReplyRabbit sends a safe prompt cache key for direct OpenAI requests when mailbox context is available. There is no OpenAI cache toggle in mailbox settings.
- Semantic retrieval: OpenAI can be selected as an embedding provider when semantic retrieval is licensed and enabled. That path can use a dedicated embedding key and is separate from using OpenAI for draft generation.
- Context: keep company context, response style, and examples current so drafts match your support policy.
- Budget: set a token budget when you need a hard monthly stop.
Use cases
- Draft customer replies in FreeScout using OpenAI as the selected cloud provider.
- Start testing on the ReplyRabbit Free tier before comparing paid provider options.
- Use OpenAI embeddings for semantic retrieval when that feature is licensed and enabled and your data policy allows provider-hosted embeddings.
Troubleshooting
| Symptom | What to check |
|---|---|
| Invalid API key error | Re-copy the key from OpenAI and confirm it belongs to the account you expect. |
| Model error | Switch to the latest option, or confirm the custom model ID is supported by OpenAI. |
| Drafts stop later in the month | Check the mailbox token budget and provider account billing, rate limits, or credit. |
| Semantic retrieval is unavailable | Confirm semantic retrieval is enabled for your plan and mailbox, then choose a supported embedding provider. |
| Draft quality is inconsistent | Improve company context, response style, examples, and the knowledge sources enabled for the mailbox. |
For broader fixes, see Troubleshooting ReplyRabbit.
Privacy and security notes
When OpenAI is selected, ReplyRabbit sends OpenAI the prompt context needed for the enabled AI feature. That can include conversation text, company context, response style rules, examples, and related support context.
Use a local provider such as Ollama if your policy requires customer content to stay on your own hardware.
Next steps
- Review the OpenAI integration for FreeScout page when you need the workflow and buying overview.
- Compare provider tradeoffs in Choosing an AI provider before changing a production mailbox.
- Continue with How to set up ReplyRabbit if company context, examples, or budgets still need attention.
FAQ
Do I need an OpenAI API key?
Yes, for OpenAI draft generation. Semantic retrieval can also use OpenAI with a dedicated embedding key when that feature is licensed and enabled.
Is OpenAI available on the ReplyRabbit Free tier?
Yes. The ReplyRabbit Free tier uses OpenAI as its AI provider. OpenAI usage is still billed by OpenAI through your provider account.
Can I choose the GPT model?
Yes. Use a configured preset, the latest option, or the Custom field for a provider-supported model ID.
Does ReplyRabbit use OpenAI prompt caching?
ReplyRabbit sends a safe prompt cache key for direct OpenAI requests when mailbox context is available. This helps OpenAI route repeated support prompts consistently, but ReplyRabbit still counts token budgets conservatively.
Does customer data go to OpenAI?
If OpenAI is selected, ReplyRabbit sends OpenAI the prompt context needed for the enabled AI feature.