AI Agents & Chat
Multi-turn conversations with memory, planning, and structured outputs.
The full AI SDK for .NET, completely free. Build production apps with agents, RAG, and local inference. Deploy commercially with no strings attached.
Multi-turn conversations with memory, planning, and structured outputs.
Built-in vector database, embeddings, and semantic search.
Connect agents to APIs, databases, and external services.
Zero cloud dependencies. Your data never leaves your infrastructure.
The NuGet package and every model in the catalog work immediately, no signup required. A Community License unlocks unlimited commercial use of LM-Kit-trained models and removes the development-mode watermark. Request yours below in 60 seconds.
> dotnet add package LM-Kit.NET
> dotnet run
using LMKit.Model;
using LMKit.TextGeneration;
var model = LM.LoadFromModelID("qwen3.5:4b");
var chat = new SingleTurnConversation(model);
Console.WriteLine(chat.Submit("Hello, world."));
No revenue caps, no team-size limits, no per-seat fees. The Community License is the SDK in full, redistributable in your own commercial product.
Side projects, startups, open-source maintainers, agencies, freelancers. Ship to paying customers under the Community License.
Internal tools, R&D prototypes, evaluation projects. No procurement step, no PO, no NDA. Install and run.
Classroom use, theses, conference workshops, academic papers. Cite LM-Kit and ship.
Embed LM-Kit.NET in software you sell. The license allows shipping the SDK as part of your product.
Start developing with LM-Kit.NET Community Edition by requesting your free license. Fill out the form below and we will verify your details and issue a free license key.
We review every request manually to keep abuse out. Expect your license key by email within one business day. No drip campaign, no sales call: just the key, plus a one-line pointer to the Quickstart.
SDK
Unrestricted access to the complete SDK. All features and components included for both development and redistribution.
Deployment
Deploy the LM-Kit.NET SDK without restriction. Ship commercial applications with confidence.
Updates
Instant access to the latest releases available to community users. Future features of the Core SDK may require a Pro license.
Models
Only models trained by LM-Kit require a Professional License. Quantized models like Llama, Phi3, and Gemma are fully available at no cost.
Support
Community Edition support is limited to bug reports. For comprehensive technical support, upgrade to the Pro Edition.
Platforms
Available on all supported operating systems: Windows, macOS, and Linux. Deploy where your data lives.
Every capability of the SDK is in the Community edition. Professional adds the support contract, the SLA, and the named-engineer relationship.
| Feature | Community | Professional |
|---|---|---|
| Licensing & Support | ||
| Instant access to latest releases | Yes | Yes |
| Commercial deployment | Yes | Yes |
| License for LM-Kit trained models | Yes | Yes |
| Technical support | Bug reports only | Full support |
| Supported platforms | Windows, Linux, macOS | Windows, Linux, macOS |
| License validity | 1 year | 1 year |
| Core AI Platform | ||
| Text generation | Yes | Yes |
| Structured output (JSON grammar) | Yes | Yes |
| CUDA, Vulkan, Metal acceleration | Yes | Yes |
| 100+ pre-configured model cards | Yes | Yes |
| Agent Orchestration | ||
| Multi-agent orchestration patterns | Yes | Yes |
| 56 built-in tools | Yes | Yes |
| 18 pre-built agent templates | Yes | Yes |
| Agent-to-agent delegation | Yes | Yes |
| Planning handlers (ReAct, CoT, ToT) | Yes | Yes |
| Skills support for conversations | Yes | Yes |
| Agents & Tool Calling | ||
| Tool calling with safety policies | Yes | Yes |
| Full MCP client (stdio & SSE transport) | Yes | Yes |
| Agent memory with context persistence | Yes | Yes |
| Human-in-the-loop controls | Yes | Yes |
| RAG & Document Intelligence | ||
| RAG with semantic search & reranking | Yes | Yes |
| DocumentRag and PdfChat | Yes | Yes |
| Layout analysis & schema discovery | Yes | Yes |
| Built-in vector DB or Qdrant | Yes | Yes |
| Vision & Multimodal | ||
| Vision & VLM support | Yes | Yes |
| Multimodal RAG (vision + text) | Yes | Yes |
| Image embeddings for multimodal search | Yes | Yes |
| VLM OCR & visual text extraction | Yes | Yes |
| Text Intelligence | ||
| Named Entity Recognition (NER) | Yes | Yes |
| PII extraction | Yes | Yes |
| Sentiment, emotion, sarcasm detection | Yes | Yes |
| Text classification | Yes | Yes |
| Translation & summarization | Yes | Yes |
| Speech & Audio | ||
| Speech-to-text transcription | Yes | Yes |
| Voice Activity Detection (VAD) | Yes | Yes |
| Hallucination suppression | Yes | Yes |
| Dictation formatting & spoken commands | Yes | Yes |
| Production Ready | ||
| Fine-tuning | Yes | Yes |
| Dynamic LoRA adapter hot-swap | Yes | Yes |
| Model quantization & optimization | Yes | Yes |
| OpenTelemetry GenAI instrumentation | Yes | Yes |
| Resilience policies (retry, circuit breaker) | Yes | Yes |
From prompts to production with zero cloud dependency.