Questions about agentic AI, Mili and your data.

Agentic AI is an approach where artificial intelligence systems perform concrete actions rather than just chatting. These agents can automate complex, multi-step tasks, interact with external systems, and make decisions based on specific goals.

Conversational AI focuses on textual or vocal exchanges. Agentic AI goes further by performing real actions: creating documents, sending emails, accessing databases, filling web forms, calling APIs. Agents act in the digital world on your behalf.

Yes. The software runs locally on your machine, so your files, credentials and client data never leave it. Inference goes over TLS to servers we operate ourselves in an ISO 27001 certified datacenter — not a US-based AI service — and the content of prompts and responses is never logged or used for training.

The software itself runs locally and keeps working on your machine. Inference — the model thinking — requires an internet connection to reach the private GPU endpoint.

The current list of supported models, with their per-token pricing, is published on the downloads page and updated as new models are added.

Mili is sold as a per-seat monthly subscription that includes a usage budget for inference; model rates are listed on the downloads page. Contact us for a quote tailored to your team.

No. You talk to Mili in plain French or English, in the browser, and it does the work. Installation is a single command; a step-by-step guide is available on the support page.

The Model Context Protocol (MCP) is an open standard for connecting AI agents to tools and data. Mili is both an MCP client and an MCP server, which is how it connects to more than 330 applications and can collaborate with other agents.

Our inference runs on servers we operate ourselves, in an ISO 27001 certified datacenter outside US jurisdiction, reached over TLS — never through a US-based AI service or a shared multi-tenant API. Only metadata (model, token counts, timestamp) is recorded; the content of prompts and responses is never stored.

Yes. Your data stays under your control on your own hardware, no prompt content is retained, a privacy officer is designated, and the privacy policy discloses exactly where inference runs and who our processors are (our own servers, never a US AI service). Our privacy policy details every category of information we process.

On GPU servers we operate ourselves in an ISO 27001 certified datacenter, outside US jurisdiction; the exact location is disclosed in our privacy policy. It is not a US-based AI service and not a shared multi-tenant API: no third-party model provider ever sees your prompt, and we keep no copy of it. Bringing that inference home to Quebec is the mission Quebec Agentique was founded on, and every architecture decision is made so that the move changes nothing for you.

macOS, Linux, Raspberry Pi and Windows. A single install command sets it up; enterprise fleet management is available for teams that deploy many agents.

Our company has two official registered names: Agentic Quebec Inc. in English and Quebec Agentique Inc. in French (no accent in the register). People also search for us as « Québec Agentique ». We are a software publisher: our product is Mili and our only website is agent-qc.net. We are not affiliated with any consulting agency, integrator or training provider using a similar name, including ones whose name is ours with the words reversed. More about Agentic Quebec Inc.