MURET, France--(BUSINESS WIRE)--Sep 15, 2026--
LM-Kit, part of the Calico IIM group, today announced the general availability of LM-Kit One, its private AI application server for Windows, Linux and macOS. The product enables organizations to build AI capabilities inside their own environment and software vendors to deploy those capabilities alongside the applications they ship.
This press release features multimedia. View the full release here: https://www.businesswire.com/news/home/20260915692425/en/
LM-Kit One combines model execution, document processing, retrieval, agents and operational controls in a single product. Teams can extract structured data from documents, build assistants that answer from internal knowledge with source references, and automate work through agents with explicit tool permissions. These capabilities are maintained and released together rather than left to each team to assemble separately.
"Organizations should not have to choose between using current AI technology and keeping control of their information," said Loïc Carrère, founder of LM-Kit. "We are building LM-Kit One for teams that need both, whether they are improving their own operations or delivering software to customers. Our investment is long term: better AI only matters when organizations can put it to work."
Control over data, models and operations
Fully local deployments run without an external AI service once models are provisioned and external tools and optional telemetry are disabled. The server does not send customer documents, prompts, results or usage telemetry to LM-Kit. Administrators control models, access policies and agent permissions, with logs and metrics retained locally for their own use.
Connections to external tools or assistants are optional. When enabled, the results of permitted tool calls may be shared with those services.
A deployment can start on one machine and expand to multiple nodes behind a load balancer, with shared databases and storage. Kubernetes is optional; virtual machines and Windows services are also supported. Applications integrate through native REST endpoints, supported OpenAI-, Anthropic- and Ollama-compatible interfaces, or the Model Context Protocol (MCP).
An established information management foundation
LM-Kit One has run within Calico IIM for more than a year, powering group software and operational workflows. The group's companies serve over 600 customers and bring more than 40 years of enterprise content management experience, including work with public institutions, education, healthcare and businesses.
LM-Kit is investing heavily in continued development of inference efficiency, document understanding, retrieval quality and agent reliability. The LM-Kit engine has received more than 150 releases since 2024, with integrated components improved and released together. The objective is to make advanced AI practical for organizations worldwide while preserving their control over data, models and deployment.
Availability and free evaluation
LM-Kit One is available to download. Evaluation and development are free for organizations of any size, with no time limit. No LM-Kit account, license key or activation is required.
Production use is also free for businesses meeting all three criteria: annual gross revenue below US$1 million, 10 employees or fewer, and no more than US$3 million in aggregate outside capital. Eligibility is assessed together with parent, subsidiary and affiliated entities. Other commercial organizations require a Professional License. Full conditions, including client projects and redistribution, are covered by the EULA. Model licenses apply separately.
Organizations and software vendors can contact LM-Kit to discuss their project, deployment requirements and commercial terms.
About LM-Kit
LM-Kit develops private AI software for documents, internal knowledge and automation. Its products are LM-Kit One, a deployable AI application server, and LM-Kit.NET, an embedded runtime for .NET applications. Based in France, LM-Kit is part of the Calico IIM group.
LM-Kit One's API reference: the Search → Answers endpoint returns grounded, citation-backed answers generated directly from an organization's own document collections.
