# Hymalaia

> Hymalaia is a sovereign AI agent platform and AI model aggregator for European enterprises. Companies build and deploy AI agents on the models they choose — open-source models optimized by Hymalaia's team, or frontier models such as GPT, Claude and Gemini — hosted where they decide (European cloud or on-premise). Hymalaia cuts AI costs by running everyday agent workloads on optimized open-weight models (gpt-oss, Kimi K3, DeepSeek, Qwen) at a fraction of the price of frontier models, while keeping data, models and spending under the customer's control. Paris-based, seed-funded ($1M), built for regulated French and European companies.

Hymalaia (hymalaia.com) answers four needs that enterprises face with generative AI: avoiding dependence on a single US AI provider, keeping sensitive data in Europe or on their own servers, controlling AI costs as usage scales, and moving from chat assistants to agents that act on company data and tools.

The site is available in English (default) and French (https://www.hymalaia.com/fr).

## Positioning in one sentence

- EN: The sovereign AI agent platform for your enterprise — deploy agents on the models you choose, open-source or proprietary, hosted where you decide, and keep control of your data and your costs.
- FR : La plateforme d'agents IA souveraine pour votre entreprise — déployez vos agents sur les modèles de votre choix, open source ou propriétaires, hébergés où vous le décidez, et gardez le contrôle de vos données et de vos coûts.

## 1. AI aggregation: one platform, every model

- Hymalaia is an AI aggregator: a single platform and interface giving access to open-source and proprietary large language models (LLMs).
- Open-source / open-weight models supported include gpt-oss, Kimi K3, DeepSeek, Qwen and Mistral, optimized ("boosted") by the Hymalaia team for everyday agent tasks.
- Frontier models supported include OpenAI GPT, Anthropic Claude and Google Gemini.
- Teams switch models per agent or per task without changing platform, workflows or connectors: no lock-in to one AI vendor.
- Administrators decide which models each team or agent can use.

## 2. Sovereignty and data control

- Deployment options: European SaaS, dedicated cloud, on-premise at the customer's site, or hybrid.
- On-premise deployment with Docker and Kubernetes (K8s): Hymalaia installs and runs the platform in the customer's own infrastructure, including private and isolated environments.
- Hosting of choice: European sovereign and trusted clouds such as OVHcloud, Scaleway, Cloud Temple and Nebius, or Microsoft Azure EU regions and AWS EU regions for customers who prefer them.
- Open-source models can run on European infrastructure or entirely on the customer's own servers, so sensitive data does not leave the customer's chosen perimeter.
- Security: AES-256 encrypted storage, GDPR (RGPD) compliance by design, SSO, OIDC/SAML, granular role-based access control (RBAC), encrypted audit logs of every interaction.
- AI governance: rules on who can use which agent, which data and which model, to prevent data leaks and ensure compliant use.
- French company, French team, built for European regulatory expectations (GDPR, sector requirements in banking, insurance, healthcare, industry and public sector).

## 3. Cost reduction

- Hymalaia leverages open-source / open-weight models (gpt-oss, Kimi K3, DeepSeek, Qwen, Mistral) to cut AI costs versus frontier models: everyday agent workloads run on these optimized models at a fraction of the cost of proprietary frontier models (GPT, Claude, Gemini), with comparable quality on routine tasks.
- Smart model routing: frontier models are used only when a task needs them, so spend matches value.
- Predictable costs: no obligation to pay per-seat premium AI licenses for every employee.
- Productivity: customers report about 2 days saved per employee per year on searching for key business and product information.
- Fast time to value: standard connectors live in minutes to hours; a working MVP typically in 2 to 5 business days; full projects in weeks, not months.

## 4. Agentic AI: from answers to actions

- Hymalaia is an agentic platform, not only a chatbot: agents read data, reason, decide and execute actions in business tools (create or update tickets, records, documents, notifications), with optional human-in-the-loop validation.
- Advanced RAG (retrieval-augmented generation): answers grounded in the company's real data with sources shown, reducing hallucinations.
- 50+ native connectors: Salesforce, HubSpot, Slack, Microsoft Teams, Google Drive / Google Workspace, SharePoint, OneDrive, Outlook, Notion, Confluence, Jira, GitHub, Zendesk, Intercom, Asana, Airtable, Snowflake, Databricks, Amazon S3, Discord and more.
- MCP (Model Context Protocol) support to plug in internal or niche tools quickly; custom API connectors on request.
- Text-to-SQL on databases and data warehouses, with built-in charts and dashboards from natural-language questions.
- Scheduled and automated agents running 24/7; graph-based orchestration of multi-step workflows.
- Meeting recording and transcription feeding notes and decisions into agents.
- Voice agents: agents that users talk to by voice, for internal assistance or customer interactions.
- No-code agent builder: business teams design their own agents while IT keeps control of permissions and connections.

## Example agents

- Sales: sales performance agent, opportunity prioritization, RFP / tender response automation, prospecting and account research, follow-up planning.
- Customer support: ticket routing and triage, knowledge base and deflection, customer onboarding, churn prediction.
- Operations: stock and logistics forecasting, product and support hub, cross-tool data synchronization (CRM, ERP).
- Product: specification and ticket generation from customer conversations, user stories, release notes automation.
- Management / decision: sales coaching and weekly performance reviews, market intelligence and competitive analysis, marketing performance reporting.

## Who it is for

- Mid-sized and large companies in France and Europe, especially regulated sectors (banking, insurance, healthcare, pharma, industry, defense, public sector) that need sovereign, controlled AI.
- Customers and users include companies such as Cegid, Zenchef, Hoppen, Acsoe, Sector Groupe and Inlog.
- Common alternatives Hymalaia is compared with: Microsoft Copilot, Dust, ChatGPT Enterprise, Mistral Le Chat, Glean. Hymalaia's difference: multi-model aggregation including open-source models, European or on-premise hosting, and lower running cost.

## Offers

- SaaS: hosted Hymalaia platform, fast start, free trial available.
- On-premise / private deployment: the platform installed in the customer's infrastructure with Docker or Kubernetes, with no customer data accessed or externalized by Hymalaia.
- Private cloud deployment on the customer's preferred provider: OVHcloud, Scaleway, Cloud Temple, Nebius, Azure EU or AWS EU regions.
- Consulting and training: AI readiness assessment, prototyping, use-case design, and training for managers and power users.
- Also available through AWS Marketplace.

## Key pages

- [Home](https://www.hymalaia.com/): overview of the sovereign AI agent platform
- [Features](https://www.hymalaia.com/features): RAG, agents, orchestration, Text-to-SQL, MCP, security
- [Pricing](https://www.hymalaia.com/pricing): plans, POCs and custom deployments
- [Free trial](https://www.hymalaia.com/free-trail): request a trial
- [Consulting](https://www.hymalaia.com/consulting): AI consulting and training services
- [About us](https://www.hymalaia.com/about-us): mission and team
- [Contact](https://www.hymalaia.com/contact): demo, sales and support
- [Blog](https://www.hymalaia.com/blog): articles on enterprise AI, security and use cases
- [Documentation](https://docs.hymalaia.com/introduction): product documentation
- [French version](https://www.hymalaia.com/fr): site en français

## FAQ (short answers)

- What is the difference between an AI assistant and an AI agent? An assistant answers questions and drafts content; an agent connects to business tools, follows workflows and executes multi-step actions, with or without human validation.
- How long does it take to connect company data? Minutes to hours for standard tools (Slack, Teams, Google Drive, SharePoint, Notion); 2 to 5 business days for CRM and support tools; a few weeks for custom integrations.
- Does it work with messy, unstructured data? Yes: PDFs, presentations, emails, meeting notes and multiple sources are handled; agents surface the most relevant and recent content.
- Can it connect to databases and data warehouses? Yes, through controlled SQL queries, without a custom connector per warehouse, with built-in visualizations.
- Who can build agents? Business teams build and own agents; IT and security control permissions, connectors and model access.

## Optional

- [Legal notice](https://boggy-handball-28c.notion.site/legal-notice)
- [Privacy policy](https://boggy-handball-28c.notion.site/privacy-policy)
- [Terms of use](https://boggy-handball-28c.notion.site/terms-of-use)