- Back to solutions
- GuardWeaver: AI agents and sovereign AI governance
GuardWeaver: AI agents and sovereign AI governance
- Product By : Matvi Consulting
Have a question about this solution?
Want a demo? Need a quote?
GuardWeaver allows companies to deploy AI agents while keeping control over what they do. The agents rely on your documents and business tools. The IT department sets the rules, the platform applies them to each action and logs everything in an audit trail. You use your own API keys, with no markup on consumption. French solution hosted on OVHcloud, available SaaS or On-Premise.
Related Documents
Related Videos
-
GuardWeaver permet aux entreprises de déployer des agents IA en gardant la main sur ce qu’ils font. Les agents s’appuient sur vos documents et vos outils métier. La DSI fixe les règles, la plateforme les applique à chaque action et trace tout dans un journal d’audit. Vous utilisez vos propres clés d’API, sans marge sur la consommation. Solution française hébergée sur OVHcloud, en SaaS ou On-Premise. GuardWeaver permet aux entreprises de déployer des agents IA en gardant la main sur ce qu’ils font. Les agents s’appuient sur vos documents et vos outils métier. La DSI fixe les règles, la plateforme les applique à chaque action et trace tout dans un journal d’audit. Vous utilisez vos propres clés d’API, sans marge sur la consommation. Solution française hébergée sur OVHcloud, en SaaS ou On-Premise.
Presentation
Deploy AI agents throughout the company without losing control of what they do.
GuardWeaver combines agent creation and governance. Your teams rely on agents connected to their documents and tools. The IT department sets the rules, and the platform enforces them for each action, in real time.
Audit log example
| Heure | Agent | Action | Décision |
|---|---|---|---|
| 10:42 | Customer Support Agent | Search in the 'After-sales Procedures' database | ✓ Allowed |
| 10:43 | Pre-sales Agent | Response to a prospect containing a client's IBAN | Masked |
| 10:45 | External HR Agent | Export of the salary grid | Blocked |
| 10:47 | Sales Agent | Update of a CRM record | ✓ Allowed |
In most companies, employees already use AI, often outside any framework. The IT department does not know which tools are circulating and what data is transiting. With the AI Act, it is now necessary to be able to show how AI is used and under what rules.
In GuardWeaver, a central agent supervises all registered agents, whether created on the platform or in another tool, and applies the same rules to them. Each action is controlled at the moment it occurs. Sensitive data is automatically masked before reaching the model. Detection of AI tools used without validation is coming soon.
On the business side, your teams create their agents without code. They respond from your documents citing their sources, deploy on your communication channels, and act within your applications via MCP and A2A protocols.
You choose your models and use your own API keys: consumption is paid directly to your providers, with no margin from us. GuardWeaver is developed in France and hosted in Europe on OVHcloud. Installation on your servers is possible upon request.
Highlights
- Sensitive data does not reach the AI : Emails, IBANs, or card numbers are automatically masked before being sent to the model.
- Even external agents follow your rules : Each action is verified before execution. An agent created in another tool, once registered, is subject to the same rules.
- Proof in case of audit : Each request and decision is recorded in an append-only log: no line can be modified or deleted.
- Sovereign, without lock-in : Hosted in Europe on OVHcloud. You use your own keys with the provider of your choice, with no margin on consumption.
Features and use cases
What the platform does, and what your teams do with it daily.
- A knowledge base that understands all your formats : Documents, audio, and video are indexed automatically, and each answer cites its source. Usage concret : Your pre-sales assistant responds to prospects on WhatsApp, Teams, or Slack with information validated by your teams.
- Workflows that know when to hand over : Conditional scenarios, out-of-hours responses, transfer to a colleague or ticket creation. Usage concret : The first level of HR or customer support is automated, complex cases remain handled by a person.
- Agents connected to your business tools : Agents connect to your applications via MCP servers and delegate tasks via the A2A protocol, each within a strictly limited scope. Usage concret : During a conversation, an agent consults a Jira ticket or updates a HubSpot record.
- Real-time governance : Usage rules in observation or blocking mode, automatic masking of sensitive data, spending caps, and an append-only audit log. Shadow AI detection coming soon. Usage concret IT, CISO, and legal management authorize AI in the company knowing exactly what the agents do.
- Costs tracked down to the user : The dashboard tracks conversations, satisfaction, and costs per model. Budget caps are set by provider, then broken down by organization, team, and user. Usage concret : The finance department manages AI adoption without bill surprises.
Features
Technical features:
| Caractéristique | Détail |
|---|---|
| Compatible models | OpenAI, Anthropic, Mistral, Gemini, OVH AI, Ollama, and any OpenAI-compatible API, with your own keys (BYOK) |
| Supported formats | PDF, Word, Excel, PowerPoint, Markdown, CSV, JSON, XML, TXT, audio, video, images, and web crawling |
| Connectors | SharePoint, Confluence, Nextcloud, Google Drive |
| Deployment channels | Web widget, WhatsApp Business, Meta Messenger, Slack, Microsoft Teams, REST API, and MCP server |
| Orchestration | Open protocols A2A (Agent2Agent) and MCP |
| Security | Hosting in Europe, workspace isolation, rules applied before each action, append-only audit log |
| Deployment | Managed SaaS, or On-Premise via Docker / Kubernetes on request |
What each plan includes
All plans provide access to AI agents and governance, regardless of the number of users.
| Inclus dans tous les plans | Détail |
|---|---|
| What AI is allowed to do | You set as many rules as you want to be alerted, and block up to 5. |
| Your sensitive data | Emails, IBANs, or card numbers are masked before being sent to AI. |
| Your AI budget | You cap spending per person, per agent, or per model. |
| Proof in case of audit | Each decision is recorded in an append-only log and kept for one year. |
A plan for every team size
Start with two or a pilot team, then upgrade as usage expands. In all plans, your agents are unlimited and governance is included.
| Plan | Utilisateurs | Idéal pour |
|---|---|---|
| XS | 2 users | A pair launching a first project, with all Premium features |
| S | 5 users | A pilot team testing a first use case |
| M | 10 users | A department, for example customer support or HR |
| L | 25 users | Several teams sharing their agents |
| XL | 50 users | An SME generalizing AI across all its departments |
| 3XL | 100 users | An ETI, or multiple sites of the same company |
AI model consumption is billed directly by your provider, with no margin from us. Beyond 100 users, we build a custom offer: just fill out the 'Contact Us' form at the top of this page.
Prerequisites, included and not included:
- Prerequisites: Your own API keys with model providers. For On-Premise deployment, Docker / Kubernetes compatible infrastructure.
- Included: Administration, multimodal indexing, omnichannel deployment, multi-agent orchestration, audit log, technical support, and SaaS updates.
- Not included: AI model consumption, billed directly by your providers. Configuration, training, and agent creation by our team, offered as additional service billed daily.
Integration with OVHcloud services
GuardWeaver relies on OVHcloud infrastructure and integrates directly with its AI services. To use OVHcloud sovereign endpoints and OVH AI models, just inject your OVHcloud API key into the platform. In SaaS or On-Premise, your data remains hosted in Europe. The REST API and MCP servers then allow connecting your agents to other OVHcloud services.
Data security and sovereignty
Hosting is European by default, and your data remains under your control whether you use the platform SaaS, On-Premise, or via OVHcloud sovereign endpoints.
- ✓ Each workspace is isolated, no cross-data access.
- ✓ Exchanges between agents are authenticated by signed tokens, with limited scopes and bounded delegation depth.
- ✓ An action outside the scope is immediately stopped, during use.
- ✓ The audit log traces every request and delegation end to end, with the decision made. It is written append-only: no line can be modified or deleted.
The solution is designed respecting GDPR and provides traceability useful for your compliance efforts under the AI Act.
Who the service is for
GuardWeaver targets SMEs and mid-sized companies, starting at about a hundred employees, who want to generalize AI without losing control of their data and costs.
Customer service, pre-sales, and HR teams use it daily to respond faster based on validated information. IT, CISO, and legal management, especially in regulated sectors, rely on governance included in each plan to authorize AI while keeping control of agent actions.
References and customer testimonials
"A performant AI conversational agent, easy to use and perfectly integrated into our digital environment. The GuardWeaver team guided us with pedagogy and pragmatism throughout the project. I recommend GuardWeaver to any company wanting to realize AI projects with a reliable, competent, and results-oriented partner."
Jonathan ClaquinGlobal Communication Director, Visionix
Visionix, a manufacturer of optical equipment present internationally, went from defining needs to production in less than a week. Testimonial collected when the platform was called RAGweaver, now GuardWeaver.
Accessing the service
From order to your first agent
- You receive a confirmation email summarizing your plan.
- The GuardWeaver team activates your access within one business day.
- You create your agents independently, no code, or entrust configuration to our team.
✓ Your first agent can be operational within the day.
Support and assistance: Technical support and updates are included in the SaaS subscription. Setup, team training, and first agent creation can be entrusted to our team as additional daily-billed service.
See GuardWeaver operate on your own use cases.
Request a demo with the 'Contact Us' form at the top of this page. We show you the platform on an example close to your business.
Prefer to start right away? The XS plan allows starting with two users, for €30 excluding VAT per month.
Support Information
Customer Service is available:
- Monday to Friday, from 9 am to 12 pm and 2 pm to 6 pm;
- by email (contact@guardweaver.com);
- in French and English.
- Conditions: support included in the subscription, request handling within 2 business days.
General Terms and Conditions of Sale
The GTC seller's terms applicable to the service can be viewed on the seller's page accessible here.