LM-Kit One is a server you install; LM-Kit.NET is a runtime you compile into your .NET application. Same engine, and one licence covers both.
The capability set is the same. What differs is where it runs, who calls it, and how it is operated.
Deploy
The private AI application server, installed on hardware you administer.
Embed
The same engine as one NuGet package, running inside your .NET process.
Six situations that bring teams to LM-Kit, and the mechanism that resolves each one.
| The problem | How it is resolved | Where |
|---|---|---|
| Invoices, contracts and forms are keyed in by hand | Structured extraction turns PDFs, scans and photos into typed fields, each with its own confidence score, so human review goes only where confidence is low. | Both products |
| Sensitive documents cannot be sent to cloud AI | Models run on your hardware. Documents, prompts and indexes stay in your environment, and outbound access is explicit and controlled. | Both products |
| Company knowledge is scattered and unsearchable | Documents are ingested into searchable collections. Hybrid semantic and full-text retrieval returns answers that cite the passages they came from. | Both products |
| Existing tools are already wired to OpenAI or Ollama APIs | LM-Kit One serves those API dialects on your network. Existing clients, coding agents and Open WebUI connect by changing a base URL. | LM-Kit One |
| AI assistants need internal access, but not free rein | Tools run behind permission policies: allow and deny rules, risk levels, approval checkpoints, and a controlled gate on web egress. MCP exposes them to assistants. | LM-Kit One |
| Your product needs AI on customer machines, sometimes offline | The runtime compiles into the application. Nothing to install beside it, no network round trip, and it keeps working disconnected. | LM-Kit.NET |
The engine underneath is identical. Each capability is reached over REST on LM-Kit One, and as a namespace in LM-Kit.NET.
Typed fields from invoices, contracts and forms, with per-field confidence. Entities and PII are located for routing or redaction.
Parse PDF, Office and email files, OCR scans, convert to Markdown, then split, merge, redact, digitally sign, and convert to validated archival PDF/A.
Multi-tenant collections with hybrid semantic and full-text search, reranking, and chat grounded in your documents with citations.
Function calling against a built-in tool catalog. Every call is checked against permission policies before it runs.
Translation, summarization, correction, sentiment, classification, keywords and language detection, with structured output under grammar constraints.
Local Whisper-family transcription with voice-activity detection, for audio and video files or live streams.
The full endpoint matrix, including what is deliberately not served, is on the compatibility page.
Eight dimensions that decide it. When rows point to different products, run both: they share the engine and the model catalog.
| LM-Kit One | LM-Kit.NET | |
|---|---|---|
| Form | A standalone server with an admin console | A NuGet library inside your application |
| Install | Signed packages for Windows, Linux and macOS | dotnet add package LM-Kit.NET |
| Who calls it | Anything that speaks HTTP: applications, teams, AI assistants | The application that embeds it |
| Interfaces | OpenAI, Anthropic and Ollama dialects, MCP, native REST | .NET API; Semantic Kernel and Microsoft.Extensions.AI bridges |
| Where it runs | Servers you administer, one node or several | Wherever the application runs, including customer machines |
| Network path | One HTTP hop inside your network | In process: no hop, works offline |
| Operations | Operator accounts, SSO, tokens, request history, audit trail | Your application's own logging and deployment |
| Upgrades | One versioned release; the whole surface moves together | One NuGet version; all capabilities move together |
One license, both products, no activation key. Nothing checks a license at runtime. Evaluation and development are free at any company size, with no time limit.
$0no key, no expiry
The complete SDK and the complete server, including commercial use and redistribution, for small companies. Evaluation and development stay free at any size.
Customannual, scaled to scope
Required above the thresholds. Scaled to deployment size, never metered by tokens, seats or end users. Carries the assurance a production deployment needs.
Get started
