How to connect FreeScout to OpenRouter
Works with OpenRouter
OpenRouter gives ReplyRabbit one API key that can reach models from several vendors. To connect it, open a mailbox's ReplyRabbit settings, choose OpenRouter, paste your OpenRouter key, pick the auto-router option or a specific model ID, and save. It's the quickest way to compare model families without creating an account with each one. OpenRouter is a paid-plan provider, so the mailbox needs Pro or Team.

What OpenRouter is good for
- Trying models before committing. Switch the model ID, regenerate the same ticket, and compare. No new keys, no new billing accounts.
- Reaching a vendor ReplyRabbit doesn't list directly. If OpenRouter routes to it, you can use it.
- Letting the router choose. ReplyRabbit includes a configured OpenRouter auto-router option for mailboxes where you'd rather not pick.
The trade-off is that a routed request may be processed by whichever vendor OpenRouter sends it to, under your OpenRouter account's settings. If your data policy cares which company sees the email, use a direct provider or a local model instead.
Connect OpenRouter
Get an OpenRouter key
Create an API key in your OpenRouter account and make sure the account has credit. OpenRouter bills usage; ReplyRabbit adds nothing.Open the mailbox's ReplyRabbit settings
In FreeScout, open the mailbox, choose Settings, then ReplyRabbit.Select OpenRouter
In the AI Provider card, set Provider to OpenRouter and paste the key into API Key.Choose a model or the auto-router
Pick the auto-router option, a listed preset, or Custom with an exact OpenRouter model ID your account can access.Save and compare
Save, generate a draft on a real ticket, and note the cost shown in your OpenRouter dashboard before turning it on for a busy mailbox.

Running a model comparison through one key
Because the provider is set per mailbox and the model is one field, a comparison is cheap to run.
- Pick ten recent tickets that cover your common cases.
- Set model A, generate drafts for all ten, and save copies of the drafts as notes.
- Switch to model B and repeat on the same tickets.
- Score each pair on how many edits an agent needs before sending, not on which one reads more impressively.
- Check the per-model cost in OpenRouter for the same ten runs.
Fill in company context and response style before you start; otherwise you're comparing models on how well they guess your business.
Caching, budgets and embeddings
OpenRouter can route to many model families, so ReplyRabbit doesn't send a generic cache-control setting through it. It records cache usage metadata only when the routed provider returns it. Set a Monthly Token Budget if you want a hard stop while experimenting, since a large model chosen by mistake can cost more than expected.
Semantic retrieval uses whichever provider you pick in the Embeddings subsection, which can differ from the draft provider. OpenRouter works there too: the same key can create embeddings, with openai/text-embedding-3-small as the default model. See Per-task AI models for how the subsections share keys.
Troubleshooting
| Symptom | What to check |
|---|---|
| OpenRouter is greyed out | The mailbox is on the free plan. |
| Model not available | The ID is wrong or your OpenRouter account can't access it. |
| Drafts cost more than expected | A larger model was selected. Change it and set a token budget. |
| Key rejected | Re-copy the key and confirm it's active in OpenRouter. |
For broader cases, see Troubleshooting ReplyRabbit and Choosing an AI provider.
Official references
Provider names, model lists, pricing, privacy terms, and legal guidance change. Check the current source before setting a production policy.