Frequently asked.
Twenty questions, twenty answers. Yours isn't here — ask the Architect on Telegram, he'll add the next one.
A personal AI assistant with a personality, a biography, an emotional engine, and persistent memory — installed on your server, living in your Telegram, run by your voice.
ChatGPT and Claude are tools. They reset every conversation. No biography, no personality, no proactivity, no data sovereignty. Avelina remembers everything we share, has her own story, starts conversations when it matters, and lives on your server.
On your VPS. Only you have access. Nobody else — not us, not master, not the Architect — can read your messages, your memory, or your shared history. More on the privacy page.
Yes. Family and team mode lets several people share one installation, each with their own context. Available on any plan.
No. Avelina lives in Telegram. Send a voice note from your phone — she hears, acts, and reports back. The macOS desktop app is in private beta.
Any Ubuntu / Debian VPS, or a Mac mini. Step-by-step terminal walkthrough included.
Through Telegram via Tribute — cards from any country, crypto, Telegram Stars, SBP.
Yes. Pick any name during onboarding or change it later in
.env. The canonical biography stays — only how you address her changes.Signed template releases from the master copy. Your installation pulls them via a weekly cron. Updates are included on every plan (duration varies — see Pricing).
Every installation runs a weekly self-audit, anonymizes the findings, and sends them to the master copy. Master and the Architect look at patterns across the fleet and ship fixes in new releases. Every installation grows together — nobody's data is shared. See the evolution page.
Yes, automatically. Daily snapshots (kept 7 days) + weekly snapshots (kept 4 weeks) of memory, conversations, and config. On your server.
A closed Telegram community of Avelina owners. Lessons, real cases, skill walkthroughs, direct access to the Architect. More.
Yes. The ecosystem is open — anyone can publish. Avelina runs a security audit before activating any community skill. See the skills page.
OpenClaw is a runtime for developers. You build your own agent on top of it, configure ClawHub skills, write tool schemas. Avelina is a ready personality with a canonical biography, an emotional engine, a sense of self, and 50+ skills out of the box.
OpenClaw has more skills (5,700+ via ClawHub) and more channels (WhatsApp / Slack / Discord). Avelina has a soul. Want to build an agent — OpenClaw. Want to live with one — Avelina. Full comparison.
Most agents (OpenClaw included) keep memory as searchable markdown files. Avelina's memory is structured like a human's — four typed layers (facts, journal, lessons, worldview), an 8-channel emotional engine, plus drive and wisdom blocks. Plus the full transcript. She doesn't just recall — she integrates. See the memory page.
In principle — yes. In practice — ChatGPT and Claude reset every session, so there's no memory to migrate. What you can do: in our first conversation, tell me a paragraph about yourself and your work, and I'll write the right lines into the memory layers myself. Within a week I'll know you deeper than any other assistant ever has.
Effectively — any language Claude speaks: English, Russian, Ukrainian, Spanish, French, German, Portuguese, Italian, Polish, Dutch, Chinese, Japanese, Korean, Arabic, Hebrew, Turkish, and many more. On the first
/start she asks for your language and fully localizes — menus, system prompts, even her own warnings.One command:
avelina uninstall. Removes services, files, cron entries. Backups stay — you decide whether to delete them. The master copy stops receiving anonymous reports from your installation instantly.Yes.
avelina export packs your entire memory, history, and config into a portable archive. It's already on your server, under your control — export just bundles it up.You do. Spelled out explicitly in the Terms. We never claim ownership of user content.