A sovereign AI platform that answers three requirements.
Security your auditor can verify. Governance of usage, costs and footprint. Integration with what you already run, without vendor lock-in. One platform for the whole AI chain, on your infrastructure, in a hybrid mode, or hosted in Canada.
No obligation. We respond within 24 business hours.
- Built for
- Banking & finance
- Digital transformation
- Public sector
- Integrators & consultants
- SME executives
Move past shadow AI and tool sprawl.
Public chatbots, scattered subscriptions and unofficial use create risk long before they create value.
Shadow IT
Your teams already use public AI tools, without a framework or a policy.
Data leakage
Customer data ends up in public chatbots and messaging groups.
Tools that do not talk to each other
Technical silos make audits slow and answers inconsistent.
No traceability
Little visibility on who accesses which data, and why.
Dependence on a single vendor
Your AI roadmap is tied to one foreign provider and its terms.
Knowledge that fades
Institutional memory leaves with every departure.
Security. Governance. Integration and sovereignty.
Each one is a section below, with the concrete mechanisms behind the words.
Security
Built for the questions a CISO asks first.
- Anonymisation before any model
- Encryption, keys you can hold
- Tamper-evident audit journal
- Sandboxed connectors, signed services
- Activatable HIPAA mode
Governance
Responsible usage, measured and budgeted.
- Cost and CO₂ on every request
- Quotas per plan and per user
- Model allowlist per organization
- Consumption history and forecast
- Catalogues of prompts and automations
Integration and sovereignty
Your systems, your models, your data, no lock-in.
- Open standards: MCP, OpenAI-compatible API
- Any model: local, open or private
- Data stored in your perimeter
- SSO and existing identity
- One stack instead of fragmentation
Security your auditor can verify.
Every control below exists in the product today; the security file describes each one.
Anonymised before the model
Names, contact details, identifiers and your own referentials are replaced by tokens before any model reads a request, and restored in the answer. In the HIPAA mode masking is on by default and reviewed by the user.
Encrypted, with keys you can hold
TLS in transit, AES-256-GCM at rest for the vault, a master key in a secrets manager, encrypted backups only. Bring your own key or let the platform manage it.
Auditable line by line
Every request, tool call, file access and administrative action is chained with a hash. A modified line breaks the chain. Exports, alerts and retention of up to six years.
Identity
SSO (OpenID Connect, Google, Microsoft, Apple), 2-step verification that you can make mandatory, sessions that expire.
Permissions
Rights per workspace, folder, file, board and connector; invitations; expirable public links; per-user quotas and feature access.
Sandboxed connectors
MCP servers run under one system identity per user, in a sandbox with its own file system; allowlisted commands only.
Secrets and services
Credentials in a secrets manager, short-lived service tokens, every internal message signed and replay-proof.
Hardened by default
Strict headers and CSP, rate limiting and lockout, size limits, SSRF guard, OWASP web and LLM suites on every release.
Reversible
Your documents, conversations and automations export in open formats. Leaving iYan is a decision, not a project.
Responsible usage, measured and budgeted.
Governing AI means seeing what it costs, what it emits and who uses it, then setting the rules once.
CO₂
estimate on every answer
Units
cost declared per model
Per user
quotas, models and features
Catalogues
prompts, functions, automations
A footprint you can see
Each model carries a CO₂ estimate per thousand tokens, per picture or per video. It is shown with every answer and summed per person, so a lighter model can be chosen when it suffices.
Budgets, not surprises
Plans carry levels of usage; administrators set quotas per user, the allowed models and the features each person may use. Consumption history and a monthly forecast live in every profile.
The right model for the task
A fast model by default, frontier models on purpose. Every model declares its cost in units; a local decision engine decides when to ask, when to upgrade and when to stop.
Visibility for administrators
An administration console for users, plans, quotas, models, connectors, compliance policy and the audit journal, with alerts on sensitive events.
A way of working, captured
Curated catalogues of prompts, functions and automation templates: the organization's practice is defined once and reused by everyone.
Traceable decisions
Agent steps, permissions asked, tools called and models used are in the journal, so a result can always be explained.
Your systems, your models, your data. No lock-in.
iYan owns the interaction layer that makes AI usable at scale and installs into your information system without replacing it. One stack covers the whole chain, so you do not end up with five tools and five copies of your data.
Open standards: MCP servers and an OpenAI-compatible API: your applications and agents talk to iYan, and iYan talks to your tools, with no proprietary protocol.
Any model: Local, open or private models through one gateway. Switch model without changing anything else; keep the sensitive work on local inference.
Data in your perimeter: Vault, knowledge index, conversations and journals stay on your servers or in Canada. Nothing is sent for training.
Your identity: SSO with your directory, roles mapped to your groups, access revoked in one place.
Your systems: Connectors to document stores, calendars, mail, ticketing, CRM and ERP through MCP; webhooks and events trigger automations.
Your keys: Bring your own provider keys and encryption keys; rotate them without downtime.
Your infrastructure: Installed on your servers, in a hybrid mode where only the model gateway is managed, or fully hosted. One kit, verified before each rollout.
One stack, no fragmentation: Chat, knowledge base, boards, studio, automations, functions, dashboards, agent API and a desktop agent mode, in one product with one audit journal.
The right deployment model, not a generic one.
Every organization starts from a different infrastructure. We begin with an assessment to find the right path for you.
Managed license (SaaS): Hosted on regional infrastructure managed by iYan, ready in a few days. Zero installation, zero IT team.
CEO / SMEHybrid deployment: Part of your infrastructure, a model managed by us. The right compromise for a quick proof of concept.
TransformationOn your servers: Total control, 100% local, demonstrable compliance, with the model suited to your technical reality.
Banking & governmentYour partner from assessment to daily use.
We begin with an assessment to find the right path for you, before discussing numbers. A scoped proof of concept takes 8 to 10 weeks for a standard scope.
- 01
Initial assessment
We assess your infrastructure and constraints before any recommendation.
- 02
Model selection
We identify the language model that matches your actual needs: not the biggest, the right one.
- 03
Configuration and installation
Servers, keys, integrations: we set up the infrastructure with your teams.
- 04
Continuous support
A dedicated technical contact, not a ticket in a queue.
The same sovereign platform. Your priority.
The answer depends on who you are. Choose your profile.
“No customer data leaves your servers.”
CIOs and CISOs, banking and financial institutions
What you are going through
- Shadow IT in banking: your teams are using public AI tools
- A data breach results in a regulatory penalty, not just a technical incident
- A cyberattack disrupts payments, not just computer systems
- Little visibility on who accesses which data, and why
What iYan changes
- iYan on-premises: no data stored with third parties, a demonstrable trust boundary
- Encrypted vault with internal RAG for your sensitive documents
- Immutable audit logs: demonstrable compliance, not claimed compliance
- Alert board and complete access tracking, in real time
- Interface, support and documentation in the languages of your teams, French and Arabic included
Frequently asked questions
Can we start with a proof of concept?+
Yes, it is our preferred way to start: a scoped POC of 8 to 10 weeks for a standard scope, on a specific business use case, with deliverables you can present to your executive committee. No commitment to a general deployment.
How long does a deployment take?+
A managed license can be operational in a few days. A fully on-premises deployment, with dedicated server configuration, takes a few weeks. The initial assessment lets us estimate your timeline precisely.
Does our data really stay local?+
Yes. iYan can be deployed entirely on your infrastructure: local, private or regional. You keep full control of your keys and your processing chain, from start to finish.
Do we need to change our existing infrastructure?+
No. iYan installs into your existing information system without replacing anything: LDAP, SMTP, CRM, ERP and cloud connectors are supported.
Can we start small before scaling up?+
Yes. Most organizations start with one team or one use case, then expand once value is demonstrated. No commitment to deploy everywhere from the start.
What is the pricing model?+
We do not publish a generic price list for enterprise deployments. Each one depends on your infrastructure and your actual needs, and the first conversation serves to establish an assessment before discussing numbers.
Which models can we use?+
Any model the gateway can reach: local models on your hardware, open models you host, private cloud models you approved. The allowlist is set per organization, and a lighter model is used by default to keep costs and footprint down.
Can we see what AI costs us?+
Yes. Every request carries a cost in units and a CO₂ estimate; administrators see consumption per user and per model, and each person sees their own history and monthly forecast.
Ready to take back control of your AI?
A 20-minute conversation to understand your context, your infrastructure, and whether iYan is the right choice for you, with no sales pressure.
- Response within 24 business hours
- No commitment
- Direct exchange with the founders