Documentation
Last updated October 25, 2026
Camelus collects feedback about your products in one Inbox, reads it, and groups similar requests into themes ranked by how many people ask and how much they matter. This page covers the first setup. For a guided tour with screenshots, from the first sign-in to telling customers something shipped, follow the tutorial.
1. Sign in and create a product
Sign in with Google at camelus.si/login. Your workspace is created on the first sign-in.
Then create a product: one of your apps or sites. Give it a name, and optionally a domain, a description and the audience it's for. The description and audience are given to the AI that classifies feedback, so a sentence or two helps. Each product gets a slug (for example habits) that identifies it in the API. The slug can't be changed later.
2. Get feedback in
Every product has these sources, each with an on/off switch in its settings:
| Source | Use it for |
|---|---|
| Feedback widget | A button on your site or app. One script tag; can identify signed-in users. See Feedback widget. |
| Email forwarding | Your support inbox. Forward it to the product's address. See Email forwarding. |
| REST API | Your backend, a support tool, Zapier or n8n: anything that can send an HTTP request. See the REST API reference. |
| Added by hand | Feedback you heard on a call, in a tweet or in a DM. Open the Inbox and click Add feedback. |
| Polar cancellations | Why customers leave. Connect your Polar organization and cancellation reasons arrive as feedback, with plan and revenue on every requester. See Polar. |
| CSV import | Old feedback from a spreadsheet or a helpdesk export. Columns are matched by name (text, email, name, created_at, external_id, title); you can change the mapping before importing. |
Items appear in the Inbox in real time, from every source and every product.
3. What happens to each message
Each item is read once by an AI model that writes an English summary, translates it if needed, and tags its type (bug, feature request, question, complaint, praise), sentiment and urgency. Personal details are kept out of the summary, and a second, rule-based pass removes any that slip through. The original text is never changed.
Similar summaries are then grouped into themes, within each product. A French and an English message about the same problem end up in the same theme. When the AI isn't sure, it asks itself a second time; when nothing matches, a new theme starts.
Messages that look like spam, or that try to give instructions to the AI, wait in a Needs review section at the top of the Inbox until you decide.
4. Themes and the score
Themes lists every theme, ranked by score. The score counts each person once (their most important message), weighs them by how they reached you, whether their identity is verified, what kind of feedback it is and how sure the AI was, and fades with time: a request loses half its weight every 30 days without news. Click the score of a theme to see exactly which people add what. AI classifies and summarizes; it never computes the score.
You're always in charge: rename a theme, move an item to another theme, create a theme from an item, merge two themes or split one. Camelus never moves an item you placed, never merges themes on its own (it suggests merges when two look alike), and lets you undo merges, splits and moves for 30 days.
5. People
People lists everyone who sent feedback with an email or a user id, across products. Merge duplicates (the same person with two emails), or delete a person on request: either their personal data only, keeping anonymized summaries for the counts, or everything.
6. Briefing, alerts and the loop
- Daily digest. Every morning at the time you pick (Settings → Notifications), Camelus emails what changed: new feedback, top and rising themes, urgent items, what shipped. The numbers come from the database; the AI only writes the two-sentence opening. On a quiet day nothing is sent. The same digest is in Briefing, with past ones. A weekly digest on Mondays is on Pro.
- Alerts (Pro) arrive right away for urgent feedback from paying customers and for themes that suddenly spike. At most 3 a day, none during quiet hours; the rest wait for the digest.
- Send to… on a theme sends it to a destination: Vicunia, n8n, Zapier, Make, Slack, Discord or your own endpoint. The app on the other side can report back, and the theme shows it under Linked work.
- Notify requesters emails the people who asked for something when it's planned, shipped or declined. The AI drafts the email; you read it and send it with one click. It only goes to people who asked to hear back, with an unsubscribe link. When a theme ships through a destination, the draft is ready for you.
- API and MCP (Pro): read themes from your code with the REST API, or let an agent ask with the MCP server.
7. Invite your team
In Settings → Team, invite a teammate by email. They open the link, sign in with that Google account, and join the workspace. Owners and admins manage products, sources, destinations, notifications, API keys and members; members read and triage the Inbox and themes, and can send themes and loop messages.
Limits
The Free plan includes 1 product, 100 feedback items a month and 1 member. Feedback that arrives past a monthly limit, or past the plan's AI budget, is still stored and shown; it's processed as soon as you upgrade, or when the month resets. It's never dropped. Upgrade in Settings → Billing; see Pricing for the plans.