HormeKeeps context. Shows the source. Asks first
Fito answers from the app's reviewed content, keeps longer conversations coherent, restores drafts and can turn some answers into reviewable actions. Understanding what you want never grants permission to change your data.
Retrieval first, answer second
The order matters: if the model answers before looking at the material, what you get is an opinion with good grammar.
It searches the material
Your question is searched against the app's guides, glossary and references with an in-house engine (BM25 with a Spanish lemmatiser). It takes about one millisecond.
It answers with that in front of it
The relevant passages are handed to the model as working material. The answer is built on them, not on whatever the model remembers from the internet.
It shows the citations
Each relevant claim carries its source, and you can tap it to open the full guide inside the app and check for yourself.
It can know your data — only if you turn it on
With the "Personal" switch on, Fito receives a snapshot of your numbers: your routine, this week's volume, your weight, your recovery and your target. With that it stops giving textbook advice and starts talking about your case.
With the switch off, it sees nothing of yours. And there is a temporary chat whose history and draft are excluded from persistence, sync, exports and backups.
Included, or bring your own key
| Fito included | Your own key (BYOK) | |
|---|---|---|
| What you need | Sign in. That is all. | A key from OpenAI, Anthropic, Google or DeepSeek. |
| Who pays | Horme, with a monthly usage quota. | You, directly to your provider. |
| Limit | A generous monthly quota; if you exceed it, your own key takes over. | Unlimited, free, forever |
| Where it travels | Your device → our proxy → the provider. | Your device → the provider. Horme never sees it. |
More than answering questions
Read your plate
One photo and Fito identifies the foods, anchors them to the local database and proposes the amounts. You confirm before anything is logged.
Listen to your day
Dictate what you ate or the set you just finished. Voice → text → log, with confirmation.
Keep longer conversations
Up to 12 context messages, with size limits so one very long answer cannot displace everything before it.
Speak out loud
Text-to-speech in neutral Latin American Spanish, with a female or male voice, for listening instead of reading.
Recover from an error
Retry resends the last question without duplicating it. Edit returns it to the composer. Your regular draft is restored when you come back.
Log wellbeing and pain
It can propose sleep, fatigue, stress, mood or a pain log. Values are bounded and sanitised, and the action always waits for your confirmation.
Explain where data travels
Included Fito uses the Horme server to verify session and quota; BYOK goes directly to the provider. The interface distinguishes both routes before sending.
Propose, never impose
Actions use a known schema, sanitisation and double validation. You see a readable preview before confirming or cancelling.
The limits, stated by us
- Fito diagnoses nothing and does not replace a health professional.
- If a question falls outside the app's material, it says so instead of inventing.
- Citations point to Horme's own content, which in turn cites peer-reviewed literature.
- An AI can be wrong. That is why it shows its source: so you can check.
Ask anything. With a source
It does not answer what sounds good: it answers what is written in the app's own material, and tells you which guide it read it in.
Ask it something hard
"Why has my bench press not moved in six weeks?" is a better question than "how many sets should I do?".
Free. No account required. Works offline from the first launch.