Frequently Asked Questions
Everything you need to know before getting started with Agora.
Platform & integration
How does integration with our software work?
We work in project mode. Integration follows 3 phases: analysis (a few hours), implementation (a few days), autonomy (a few weeks). Our dedicated project team (1 to 2 people on the Agora side) connects to your APIs and configures the agents. On the publisher side, you typically need to involve 1 to 2 people during the implementation phase.
What team size is needed on the publisher side?
1 to 2 technical people during the integration phase (a few days). After that, the no-code studio allows non-technical profiles to manage agents on a daily basis.
Is it compatible with our tech stack?
Integration is done via REST API. If your software exposes APIs, we can connect to them. We also support webhooks and real-time data streams.
Pricing
How does your pricing model work?
Our pricing is not public as it is tailored to each project. The model is based on a monthly flat rate. No per-token billing or unpredictable variable costs. Contact us to discuss your project and get a proposal.
Are there hidden costs (tokens, APIs, etc.)?
No. LLM inference runs on our GPU (or yours in edge mode). No per-token billing, no external API costs. The flat rate includes inference, hosting, and support.
Sovereignty & security
Where is the data hosted?
On our GPU servers in France. No data passes through US servers or US cloud APIs. Optionally, we can deploy the stack directly on your infrastructure (edge mode).
Are you GDPR compliant?
Yes, by design. Hosted in France, local processing, no subcontracting to US clouds. DPA available on request.
Can we deploy on-premise?
Yes. Our inference stack (LLM + vector database) can be deployed on your GPU. We handle deployment, configuration, and ongoing maintenance.
How do you handle data security?
Security is built in from the ground up. Strict per-tenant data isolation, encryption in transit and at rest, complete audit logs. Your users' data is never exposed to third-party LLMs. The founding team has strong cybersecurity experience, and these principles have been embedded in the product architecture since day one.
How do you ensure agent quality over time?
Every project includes a continuous quality process. We implement non-regression testing, scope controls, and drift detection to ensure agents stay within the defined boundaries. Behaviors are traced and measured, allowing us to identify any unexpected changes and address them before they impact your users.
Technology
Which LLM models do you use?
We deploy the best open-weight models: Qwen, Mistral, Llama, DeepSeek. The choice depends on your use case. We handle multi-model orchestration automatically.
Does it work in multiple languages?
Yes. The platform supports multiple languages with automatic language detection. Speech recognition adapts to the user's language without any configuration.
What's the difference with a classic chatbot?
A classic chatbot answers questions with pre-written responses. Our agents understand intents, execute actions in your software, generate content, search documents, and remember context across sessions. It's an operational assistant, not an automated FAQ.