Capability · Residency
Stored in Switzerland. Processed in the EU.
Persistent data stored in Zurich, Switzerland; processing on EU infrastructure; and a Swiss model track that keeps inference in Switzerland when you choose it.
Data residency claims in AI are usually vague on purpose. monopea’s is specific because it can afford to be: your persistent data — conversations, memory, knowledge graph, documents, encrypted secrets — is stored at rest in Zurich, Switzerland, and processing runs on EU infrastructure.
And when residency matters for inference too, the Swiss model track keeps it in-country: Infomaniak-hosted open-weight models running in Switzerland, selectable as your default. No slogans about data “never leaving” anywhere — just a precise map of where each thing happens, so you can make the call.
Where your data is stored
Everything monopea persists for you — conversation history, long-term memory, the knowledge graph, uploaded documents, and envelope-encrypted secrets — is stored in Zurich, Switzerland, on Supabase-managed Postgres in the AWS eu-central-2 region. That is a statement about data at rest: the durable record of your work lives in Switzerland.
Storage residency is the part that endures. Requests come and go; the accumulating asset — the memory and knowledge your agent builds over months — sits under Swiss jurisdiction, isolated per tenant by Postgres row-level security.
Where processing happens
Processing is European, and we say so plainly: the agent runtime runs on EU infrastructure (Fly.io, Paris), and the dashboard and API edge run on Cloudflare. In-flight work is handled in the EU, then written back to storage in Zurich. Stored in Switzerland, processed in the EU — that is the honest shape of it.
Model choice is equally explicit: if you point the agent at an external frontier model — Claude, GPT, Gemini — that inference call goes to the provider you chose, under its terms. Nothing else about where your data is stored or where the runtime operates changes.
The Swiss model track: inference in Switzerland
Choose the CH model track and inference itself stays in Switzerland: Qwen3.5-397B on the high tier, Mistral 24B on the mid tier, and Nemotron Nano 30B on the low tier — open-weight models hosted by Infomaniak in Swiss data centres. Your prompts are processed by a Swiss host, not shipped to a US API.
This is the proof behind “any model”: the governance layer is model-neutral, so a Swiss-hosted open-weight default and an external frontier model run under the same gate, the same vault, and the same audit trail. You trade nothing in control by choosing either.
- CH track tiers: Qwen3.5-397B · Mistral 24B · Nemotron Nano 30B — hosted by Infomaniak
- External models — Claude, GPT, Gemini, and more — remain selectable under the same governance
- Switching tracks changes where inference runs, not how it is governed
Residency that follows every call, not just chat
Region routing is re-resolved for every model call the platform makes on your behalf — background summarisation, embeddings for search and memory, and internal evaluations included — not only the chat turn in front of you. A CH-track tenant’s side-channel work resolves to the same Swiss endpoints as its conversations.
This is the detail most residency stories skip. A platform can route your chat to a compliant endpoint and still embed your documents through whatever is cheapest. monopea resolves the region centrally, and every call type inherits it.
Why it matters
A precise answer to “where is my data?”
Zurich for storage, EU for processing, Switzerland for CH-track inference. Specific enough for your own privacy policy.
Swiss inference when you choose it
Infomaniak-hosted open-weight models keep prompts and completions in Switzerland on the CH track.
No residency leaks in the side channels
Summaries, embeddings, and background work re-resolve to the same region as chat.
FAQ
Swiss residency & routing, in short
- Where is my data stored?
- Persistent data — conversations, memory, knowledge graph, documents, encrypted secrets — is stored in Zurich, Switzerland (Supabase Postgres on AWS eu-central-2). Processing runs on EU infrastructure: the agent runtime in Paris and the dashboard edge on Cloudflare.
- Does my data ever leave Switzerland?
- At rest, no — storage is in Zurich. In flight, processing happens on EU infrastructure, and if you select an external model such as Claude or GPT, that inference goes to the provider you chose. On the CH model track, inference stays in Switzerland with Infomaniak. We state the split precisely instead of claiming “never leaves”.
- Which models run in Switzerland?
- The CH track runs Infomaniak-hosted open-weight models: Qwen3.5-397B (high tier), Mistral 24B (mid), and Nemotron Nano 30B (low). Choose the track as your default and inference resolves to Swiss endpoints.
- Is Monopea certified under GDPR, FADP, or ISO 27001?
- No formal certification has been obtained yet. The architecture is built for compliance-grade controls — Swiss storage, EU processing, tenant RLS isolation, signed audit trails — and we describe it exactly rather than borrowing a badge.
- Does residency apply to embeddings and summaries?
- Yes. Region routing is re-resolved for every call type — embeddings, background summarisation, and internal evaluations — not just the visible chat turn.
Keep exploring
Secrets vault
A write-only vault: the agent references {{secret:NAME}}, the plaintext is substituted only at dispatch, and the model never sees the value.
Approval gate
Default-deny control: mutation-capable tool calls become pending proposals and wait for your decision before anything is dispatched.
Long-term memory
A private, persistent memory and knowledge graph — tenant-isolated, stored in Zurich, and never used to train shared models.