A smart drive that reads, understands and connects your documents. But DEPOSIUM® goes further: its agentic layer enables cross-document, consolidated analysis — without hallucinations. It's a RAG (Retrieval-Augmented Generation) designed to respect planetary boundaries — CSR by design, digital sobriety, zero fabrication.

Generative AI consumes enormous amounts of energy. DEPOSIUM® takes the opposite approach: compact models (< 500 MB), enrichment at ingestion rather than at each query. Our ET(K)L architecture drastically reduces the number of LLM calls needed — no reprocessing, no waste. Digital sobriety by design.

Pricing & Plans

Q: Why not just use the OpenAI API directly?

You could — just like you could host your own server to sell online. But you use Shopify. OpenAI sells building blocks: tokens, a model, an API. To actually query your documents, you'd still need to build the ingestion pipeline, the vector database, the search engine, the hallucination guardrails, the hosting, the GDPR compliance… and several months of development. DEPOSIUM® is a platform: you drop your files, you ask your questions, you get sourced answers. That's it.

Q: Can I try it for free?

Yes. The free trial gives you access to the platform for 7 days with 100 pages and 100 queries, and 10 MB of storage — no commitment, no credit card. API and MCP included. Your quotas are fungible (1 page = 4 queries). If you upgrade to a paid plan, your remaining quotas are transferred.

Q: Where is my data located?

Our server infrastructure and databases are hosted in Amsterdam (Netherlands, EU) by Railway. Inference relies on a pool of operators running open-weight models in Europe — Groq (Helsinki datacenter), OVHcloud (France), Fireworks AI (EU region) and Mistral AI (France) — on certified infrastructure (SOC 2 Type II, ISO 27001, GDPR). As some operators are US-incorporated, any transfers are framed by the EU-US Data Privacy Framework and/or Standard Contractual Clauses. Embeddings are hosted on our own infrastructure — only the data needed for inference transits. Fully auditable ingestion pipeline. For complete isolation, we offer On-Premise deployment.

Q: Which LLMs are used?

DEPOSIUM® uses open-weight models run in Europe by a pool of inference operators — Groq (EU datacenter in Helsinki), OVHcloud (France), Fireworks AI (EU region) and Mistral AI (France) — on certified, GDPR-compliant infrastructure (SOC 2 Type II, ISO 27001). As some operators are US-incorporated, any transfers are framed by the EU-US Data Privacy Framework and/or Standard Contractual Clauses. Embeddings run on our own infrastructure — only the data needed for inference transits to LLM services. Your documents are never used to train models.

Q: Is there a commitment?

No commitment at all. Billing is monthly and you can cancel anytime. Your data remains exportable.

Q: How does DEPOSIUM® prevent hallucinations?

Every answer is generated from your documents, never made up. DEPOSIUM® enriches your data at ingestion (entity extraction, knowledge graph, structural analysis) so the AI has rich, verified context. On our benchmark (854 pages, maintenance manual), we measure an F1 Score of 96% — meaning 96% of relevant information is correctly retrieved — with 100% precision: zero erroneous facts introduced.

Q: What document types can I import?

Currently, DEPOSIUM® supports PDF import (with built-in OCR for scanned documents). Markdown, plain text, PowerPoint and Word (doc/docx) formats are coming soon.

Q: Can I connect my existing tools?

DEPOSIUM® offers an MCP server and CLI access to integrate into your workflows. Native connectors (Drive, SharePoint, Notion...) are coming soon — feel free to contact us to let us know which connector you'd like to see prioritized.

EU hosting. GDPR compliant.

Free trial

€0

Credits never expire

Try the full pipeline, no commitment.

100 pages · 100 queries included
10 MB storage
1 user
Web + MCP
No top-up

Remaining quotas transferred when upgrading.

Get started

Starter

€47 €47 excl. tax /month

i.e. €468 excl. tax/year

Taxes based on your country

Your everyday intelligent RAG.

200 pages · 200 queries/month
500 MB storage
1 user
Web + MCP

Get started

Pro

€139 €139 excl. tax /month

i.e. €1,308 excl. tax/year

Taxes based on your country

Automation and precision at scale.

500 pages · 500 queries/month
2 GB storage
1 user
Web + MCP + CLI

Launch offer: first 100 users — 30 min onboarding included

Get started

Teams

€299 excl. tax /month

i.e. €3,588 excl. tax/year

Taxes based on your country

Your team, your data, your roles.

3,000 pages · 3,000 queries/month
10 GB storage
10 users (+€15/extra user)
Web + MCP + CLI + RBAC
1h onboarding

Get started

Enterprise / On-Premise

For large enterprises and regulated industries.

Custom pages & queries
Dedicated or On-Premise infrastructure
SSO + Audit logs
Web + RBAC + CLI + API + MCP
24/7 support

Developers & API

MCP

MCP Server

Connect DEPOSIUM® to Claude, Cursor, or any compatible AI agent.

CLI

Automate your ingestion workflows from the command line. Available from Pro plan.

SSE

SSE Streaming

Real-time responses, token by token.

Datalake

Plug your ingestion pipeline into the CLI.

1 CLI call = 1 page · 1 MCP call = 1 query

Cost simulator

Set one value, the others adjust automatically.

Budget

€

Pages

Queries

req.

← pages 50 / 50 Queries →

Standard rate

1 page = 4 queries €0.12/page · €0.03/query

Fungible quotas — Pages ↔ queries: your quotas are interchangeable. Never blocked on one type as long as you have quota left on the other.

Unlimited workspaces · Unlimited history (counted in storage) · Monthly rollover, valid 1 year

Pay-as-you-go top-up

Anytime, no commitment. Choose your pages/queries split.

Amount	Price/page	Price/query	Discount
< €50	€0.12	€0.03	Standard rate
€50 – €199	€0.11	€0.027	-10%
€200+	€0.10	€0.024	-20%

Supported languages

Fully supported English French German Spanish Italian Portuguese

Likely supported — not guaranteed

Dutch Swedish Finnish Danish Hungarian Romanian Russian Turkish Arabic

Partial support — not guaranteed, degraded mode possible

Chinese (Mandarin) Japanese Polish Indonesian Vietnamese Korean Persian Czech Norwegian Ukrainian Thai Greek Hindi Bengali Malay Slovak Croatian Serbian Slovenian Catalan Chechen

All languages not listed above are not supported.

Compare plans in detail

Full pipeline from day one. 7-branch hybrid search, anti-hallucination knowledge graph, precision reranker and Guardian validation — identical on every plan. What changes is capacity.

	Free trial	Starter	Pro	Teams	Enterprise / On-Premise
Monthly quotas
Pages/month	100	200	500	3 000	Custom
Queries/month	100	200	500	3 000	Custom
Storage	10 Mo	500 Mo	2 Go	10 Go	Custom
Fungible quotas	✓	✓	✓	✓	—
Top-up	—	✓	✓	✓	—
Search pipeline
BM25 + text search	✓	✓	✓	✓	✓
Vector search (3 branches)	✓	✓	✓	✓	✓
Knowledge Graph	✓	✓	✓	✓	✓
Precision reranker	✓	✓	✓	✓	✓
Guardian (post-generation validation)	✓	✓	✓	✓	✓
Performance
Requests/min (rate limit)	20	60	120	300	Custom
Max results per search (top_k)	10	20	50	100	Custom
Max scope per query	100	5K	20K	100K	Custom
Timeout per request	15s	30s	45s	90s	Custom
Access
Web UI	✓	✓	✓	✓	✓
REST API	✓	✓	✓	✓	✓
MCP Server	✓	✓	✓	✓	✓
CLI	—	—	✓	✓	✓
Collaboration
Users	1	1	1	10 (+)	Unlimited
Spaces	Unlimited	Unlimited	Unlimited	Unlimited	Unlimited
RBAC (roles, spaces)	—	—	—	✓	✓
SSO	—	—	—	—	✓
Audit logs	—	—	—	—	✓
Support & SLA
Community Discord	✓	✓	✓	✓	✓
Email support	—	—	✓	✓	✓
Onboarding	—	—	—	1h	Dedicated
24/7 support	—	—	—	—	✓

Frequently asked questions

What is DEPOSIUM®?

Why not just use the OpenAI API directly?

You could — just like you could host your own server to sell online. But you use Shopify. OpenAI sells building blocks: tokens, a model, an API. To actually query your documents, you'd still need to build the ingestion pipeline, the vector database, the search engine, the hallucination guardrails, the hosting, the GDPR compliance… and several months of development. DEPOSIUM® is a platform: you drop your files, you ask your questions, you get sourced answers. That's it.

Can I try it for free?

Where is my data located?

Our server infrastructure and databases are hosted in Amsterdam (Netherlands, EU) by Railway. Inference relies on a pool of operators running open-weight models in Europe — Groq (Helsinki datacenter), OVHcloud (France), Fireworks AI (EU region) and Mistral AI (France) — on certified infrastructure (SOC 2 Type II, ISO 27001, GDPR). As some operators are US-incorporated, any transfers are framed by the EU-US Data Privacy Framework and/or Standard Contractual Clauses. Embeddings are hosted on our own infrastructure — only the data needed for inference transits. Fully auditable ingestion pipeline. For complete isolation, we offer On-Premise deployment.

Which LLMs are used?

DEPOSIUM® uses open-weight models run in Europe by a pool of inference operators — Groq (EU datacenter in Helsinki), OVHcloud (France), Fireworks AI (EU region) and Mistral AI (France) — on certified, GDPR-compliant infrastructure (SOC 2 Type II, ISO 27001). As some operators are US-incorporated, any transfers are framed by the EU-US Data Privacy Framework and/or Standard Contractual Clauses. Embeddings run on our own infrastructure — only the data needed for inference transits to LLM services. Your documents are never used to train models.

Is there a commitment?

Is AI polluting?

How does DEPOSIUM® prevent hallucinations?

Every answer is generated from your documents, never made up. DEPOSIUM® enriches your data at ingestion (entity extraction, knowledge graph, structural analysis) so the AI has rich, verified context. On our benchmark (854 pages, maintenance manual), we measure an F1 Score of 96% — meaning 96% of relevant information is correctly retrieved — with 100% precision: zero erroneous facts introduced.

What document types can I import?

Can I connect my existing tools?