A private AI agent for the whole team, DPA included.
A Business workspace gives your team Evanesca's private AI agent and chat — and its API — with what your DPO will ask for: a signed DPA (Article 28 of the GDPR), the list of sub-processors, every model run in Europe, SSO and roles, spending limits on top of each task's cap, VAT invoices and passport exports.
Checked on 29 Sep 2026 · Mistral AI, Infomaniak and Linkup pages read on 13 Oct 2026
Opening soon: what follows describes the service as it will work.
What a Business workspace adds
Tasks, capped one by one
Members hand the agent tasks — sourced research, Word, PDF and Excel documents, images — and each task has its own spending cap (€0.50, €1, €2, €5 or €10), on top of the workspace's limits. It never runs code. Agent mode.
A signed DPA
A data processing agreement under Article 28 of the GDPR, signed between your company and FLEECE AI, which publishes Evanesca.
Named, with notice
The list of sub-processors comes with the DPA. Before any change, you get notice and can object. Current list.
Your team's messages never leave Europe
Every model Evanesca serves runs in Europe — in France, in Mistral AI's EU region, in Germany or in Switzerland — for every member, Agent mode included. No model call leaves Europe: not as a backup, not when servers are busy. Web search, off by default, is the one step whose servers Evanesca has not verified (see below).
SSO and roles
Single sign-on through your identity provider, and roles for members and admins.
Spending limits
Set spending limits so the team's usage stays within the budget you choose, on top of the cap of each Agent task.
VAT invoices
Invoices that show VAT, for your accounts.
Passport exports
Export the passports of your team's answers — model, operator, country, applicable law, what was kept — for your records. Without the content of the messages.
No copy of your messages
Prompts and answers stay in memory only while the answer is written; each member's history — and the files the agent writes — live in their own browser.
Named providers
Web search, off by default and billed per call (€0.0076 a search, €0.0015 a page), goes through Linkup, a French company whose registered office is in Fresnes: the model's queries and the pages it reads, never the conversation. Images are made by Infomaniak in Switzerland, about €0.02 each. Web search · Images.
What happens to your team's data
The answers to give your DPO, point by point. Each route is detailed on the passport.
- Prompts and answers
- Never stored by Evanesca: in memory while the answer is written, then forgotten.
- History
- In each member's browser, not on Evanesca's servers.
- What Evanesca keeps
- For each message: the tokens and the cost, for billing. For each Agent task: a record without content (model, cap, status, dates). Account and payment data: as long as the law requires.
- Models
- All run in Europe — in France by Scaleway (Paris) and OVHcloud (Gravelines, as a fallback), in Mistral AI's EU region, in Germany by AKI.IO or in Switzerland by Infomaniak (outside the EU, under an EU adequacy decision) — by European companies: OVHcloud, Scaleway, Mistral AI and Infomaniak have no US parent, out of the CLOUD Act's direct reach; AKI.IO does not publish its shareholders. Evanesca serves no other model.
- What the operators keep
- OVHcloud states “zero data retention: we keep only the data required for billing purposes” (OVHcloud AI Endpoints). Scaleway may keep and read requests that cause errors or look abusive, for up to 14 days (Scaleway's data privacy page). Mistral AI keeps API inputs and outputs “for thirty (30) rolling days to monitor abuse” (Mistral's privacy policy). AKI.IO states that “prompts, inputs, and generated outputs are processed in volatile memory and are not logged, stored, or analyzed by AKI.IO” (AKI.IO); its privacy policy keeps usage data — request ID, model, status, timestamps, token usage — for 90 days, for billing and service maintenance (AKI.IO's privacy policy). Infomaniak states that queries are “neither recorded nor used to train models” (Infomaniak AI Tools).
- Agent mode
- The loop runs on Evanesca's server, at Vercel (Paris region); each model call goes to its model's operator, in Europe; the agent's files stay in the member's browser, and the server holds them in memory only while a task runs.
- Web search
- Off by default, in the chat and in Agent mode. When a member switches it on, Linkup Technologies, a French company (registered office: Fresnes), receives the queries the model writes and the address of the pages to read — never the conversation. Evanesca has not verified where its servers run; its own privacy policy applies. Billed per call: €0.0076 a search, €0.0015 a page read.
- Images
- Made by Infomaniak in Switzerland, then sent back to the member's browser. A face photo is personal data: members send only photos they have the right to use.
- Dictation
- Transcribed by Mistral AI, a French company. Mistral AI keeps API inputs and outputs “for thirty (30) rolling days to monitor abuse” (Mistral's privacy policy). Mistral AI lets its customers object to training on that data, and Evanesca's account objects.
- Training
- Never by Evanesca. OVHcloud and Scaleway both state that they don't train on customer data, Infomaniak states that queries are not used to train models, and AKI.IO that customer prompts and outputs are not used to train AI models; Mistral AI lets its customers object to training, and Evanesca's account objects, except for what is sent to Mistral Large 4: it is a preview, and Mistral's commercial terms let it use the data of preview models for training, whatever the objection.
Setting up a workspace
- Write to usAn email to contact@fleeceai.agency: your company's name, and what your team will do with Evanesca.
- Review the DPAWe send the DPA and the current list of sub-processors. Your DPO reviews them before anything is signed.
- Set the rulesRoles and spending limits: your team starts with the rules already in place.
A question before you write? Every way to reach us.
Frequently asked questions
Do you sign a DPA?
Yes. A Business workspace comes with a data processing agreement under Article 28 of the GDPR, signed, and the list of sub-processors.
Can a member send company data outside Europe?
Not to a model: every model Evanesca serves runs in Europe, for every member, Agent mode included. Web search is off by default; if a member switches it on, the queries the model writes go to Linkup, a French company — Evanesca has not verified where its servers run.
Is our data used to train AI?
Never at Evanesca. OVHcloud and Scaleway both state that they don't train on customer data, Infomaniak states that queries are not used to train models, and AKI.IO that customer prompts and outputs are not used to train AI models. Mistral AI, which runs Mistral Large 4, Mistral Small 4 and dictation, lets its customers object to training, and Evanesca's account objects, except for what is sent to Mistral Large 4: it is a preview, and Mistral's commercial terms let it use the data of preview models for training, whatever the objection.
Can we cap what the agent spends?
Yes. Every Agent task has its own cap — €0.50, €1, €2, €5 or €10 — never above the balance, web calls and images included, and the workspace's spending limits apply on top.
Where do the agent's files go?
Into the member's browser, with the conversation. Evanesca's server holds them in memory only while a task runs; download them to share or archive them.
Where is the list of sub-processors?
On the legal notice, with the operators of every model. Before any change, Business workspaces get notice and can object.
Do we keep a record of what the team asked?
Not at Evanesca: there is no copy of prompts or answers. Each member's history stays in their browser, and passport exports record which model answered, where and under which law — without the content.