Built in Europe. For humans, not platforms.

Your AI. Your Memory. Period.

SYNAE.ai is an assistant with user-owned memory: add/edit/export in one click, transparent provenance, switchable dialogue modes — and a lower ecological footprint.

Try the MVP Read the docs Download one-pager (PDF)

Value & Use Cases

Explicit memory

You decide what remains: add, edit, export in one click.

Visible provenance

Clear indicators of source, latency and estimated costs.

Dialogue modes

Assistant, muse, Socratic, contrarian, dreamer.

Multi-thread & JSON export

Work like you think. Export journals and memory cards.

FR/EN UI

Seamless bilingual interface.

Quick use cases

Why it matters

🛡️

Data sovereignty

User-owned memory = individuals control what is kept, edited or purged. This framing resonates with academics and policymakers.

Ownership • Portability • Control
🔎

Trust layer (provenance-first)

Transparent provenance counters disinformation: show sources, latency and system path so results are auditable.

Traceability • Auditability
🌱

Eco-design (measurable)

Selective memory and RAG-lite reduce compute and bandwidth. We’ll publish lightweight metrics (median latency, avoided tokens/transfers).

Less compute • Less transfer
🇪🇺

Future-proof (EU alignment)

Proactive alignment with EU frameworks (Data Act / AI Act spirit) signals risk-aware design that funders and enterprises value.

Compliance by design

How it works

Memory cards

Tags, inline edit, import/export.

Relevant injection

RAG-lite/top-K retrieval (incoming) to avoid over-stuffed prompts.

Provenance journal

Latency, backend source, cost hints — all visible.

Auto-persistence

Local store.json for memory/logs.

Stack

FastAPI backend + lightweight Web UI.

Lower footprint, higher signal.

Fewer calls, fewer tokens

Selective memory avoids re-sending context. Only what’s needed is retrieved.

  • Expected impact: fewer redundant API calls
  • Smaller prompts; lower bandwidth

Local-first options

When relevant, processing & storage can stay on the device (exports, caches).

  • User controls persistence & deletion
  • Portability by design (JSON exports)

Relevance over brute force

RAG-lite/top-K (incoming) to retrieve minimal, high-signal context.

  • Less compute churn; stable quality
  • Visible metrics (latency, size)

We’ll publish a lightweight ‘metrics’ page: median latency, context size, and avoided transfers.

Security, Transparency & Sovereignty

For individuals & families

Parental controls

Optional content filters and usage insights. Parents can supervise without reading private messages.

Family accounts

Profiles per family member; shared rules (time windows, mode limits), separate memories.

Consent & transparency

Clear logs of what’s stored or deleted, with export at any time.

Démo & captures (MVP)

Chat threads with memory cards
Fils de discussion avec cartes mémoire
Memory cards list and inline edit
Cartes mémoire, édition inline
Provenance journal
Journal de provenance
Origin (metrics view)
Vue métriques de provenance

Captures du prototype MVP (v0.x). Remplacez les images dans /screens/ si vous mettez l’UI à jour.

Roadmap

Who we are & Contact

SYNAE.ai is a European project focused on ethics, transparency and frugality.

FAQ

Puis-je exporter toute ma mémoire ?

Oui — export complet en JSON.

Qui voit mes cartes ?

Vous seul(e), sauf partage explicite.

Un fournisseur précis est-il obligatoire ?

Non. Un mode dégradé de secours est prévu.

Quid des coûts ?

Journal & estimations de tokens si une clé est fournie.

Puis-je définir un contrôle parental ou des règles familiales ?

Oui. Contrôle parental optionnel, plages horaires et règles partagées sont prévus pour les comptes familiaux.

Les parents lisent-ils les messages de l’enfant ?

Non. La supervision repose sur des réglages et des indicateurs d’usage, pas sur le contenu des messages.