Your AI. Your Data. On Your Device.
LM-Kit.NET is the only full-stack AI framework for .NET that unifies everything you need to build and deploy AI agents with zero cloud dependency. It combines the fastest .NET inference engine, production-ready trained models, agent orchestration, RAG pipelines, and MCP-compatible tool calling in a single in-process SDK for C# and VB.NET.
🔒 100% Local · ⚡ No Signup · 🌐 Cross-Platform · 📦 Zero Dependencies
Get started with the LM-Kit Community Edition today. Whether you're a hobbyist, startup, or open-source developer, the Community Edition provides full access to build and experiment.
👉 Claim Your Free License Now!
LM-Kit.NET is a highly technical SDK that brings cutting-edge AI research directly into the .NET ecosystem. Our engineering team continuously ships the latest advances in generative AI, symbolic AI, and NLP - with weekly releases that add new model architectures, optimize inference pipelines, and expand capabilities across the entire stack.
Check our changelog to see the pace of innovation.
Listed from most recent to oldest
- 🛰️ Agentic Framework - Build autonomous AI agents with MCP support, 56 built-in tools, 18 agent templates, multi-agent orchestration, planning strategies, Agent Skills Protocol, and resilience policies - all running locally
- 🔧 Complete MCP Client - Full Model Context Protocol implementation with HTTP/SSE and Stdio transports, supporting tools, resources, prompts, sampling, roots, elicitation, progress tracking, cancellation, logging, and subscriptions
- 🤖 Agent Orchestration - Compose multi-agent workflows with Pipeline, Parallel, Router, and Supervisor orchestrators
- 🧠 Planning Strategies - ReAct, Chain of Thought, Tree of Thought, Plan and Execute, and Reflection handlers for sophisticated reasoning
- 🎯 56 Built-in Tools - Production-ready tools for data, text, numeric, security, utility, I/O, and network operations
- 📋 18 Agent Templates - Pre-configured specialists including Chat, Assistant, Tool, ReAct, Code, Writer, Analyst, Planner, Research, Reviewer, Summarizer, Extractor, Tutor, Translator, Classifier, Debugger, Editor, and QA agents
- 🎓 Agent Skills Protocol - Reusable task specialists with SKILL.md specification, progressive loading, and semantic matching
- 📄 PDF Chat + Document RAG - Import PDFs, index them locally, and chat with them using the new
DocumentRagandPdfChatAPIs (with built-in PDF attachments, chunking, and a local vector store) - 🔧 Tool Calling for Local Agents - Build AI agents with state-of-the-art tool calling. Supports all modes (simple, multiple, parallel) with structured JSON schemas, safety policies, and human-in-the-loop controls
- 🎙️ Speech-to-Text - Convert spoken audio into highly accurate text transcripts with voice activity detection, supporting 100+ languages
- 👁️ VLM-Based OCR - High-accuracy text extraction from images and scanned documents using vision language models
- 🛡️ Multimodal PII Extraction - Identify and extract personally identifiable information from text and images for compliance
- 🏷️ Multimodal Named Entity Recognition - Detect and classify entities (people, organizations, locations, etc.) across text and images
- 🌐 Multimodal RAG with Reranking - Improve accuracy with multimodal retrieval-augmented generation and semantic reranking
- 🧬 Built-in Vector Database Engine - Store and retrieve embeddings at any scale without external dependencies
- 🔗 Vector Database Connectors (Open Source) - Integrate with Qdrant for semantic search and hybrid RAG pipelines
- 🧠 Semantic Kernel Integration (Open Source) - Build intelligent workflows with Microsoft's Semantic Kernel + LM-Kit.NET
- 👁️ Vision Support - Image understanding with vision language models
- 🎮 Vulkan Backend - Accelerated multi-GPU support for AMD, Intel, and NVIDIA
- ✨ Dynamic Sampling - Up to 75% error reduction and 2x faster processing
A complete AI stack with no moving parts. LM-Kit.NET integrates inference, models, orchestration, and RAG into your .NET application as a single NuGet package. No Python runtimes, no containers, no external services. Everything runs in-process.
Not every problem requires a massive LLM. Dedicated task agents deliver faster execution, lower costs, and higher accuracy for specific workflows - with complete data control and minimal resource usage.
| Benefit | Description |
|---|---|
| Complete Data Sovereignty | Sensitive information stays within your infrastructure |
| Zero Network Latency | Responses as fast as your hardware allows |
| No Per-Token Costs | Unlimited inference once deployed |
| Offline Operation | Works without internet connectivity |
| Regulatory Compliance | Meets GDPR, HIPAA, and data residency requirements by design |
- Autonomous AI agents that reason, plan, and execute multi-step tasks using your application's tools and APIs
- Multi-agent workflows with Pipeline, Parallel, Router, and Supervisor orchestration patterns
- MCP-connected agents that access thousands of community MCP servers while keeping inference local
- RAG-powered knowledge assistants over local documents, databases, and enterprise data sources
- PDF chat and document Q&A with retrieval, reranking, and grounded generation
- Voice-driven assistants with speech-to-text, reasoning, and function calling
- OCR and extraction pipelines for invoices, forms, IDs, emails, and scanned documents
- Compliance-focused text intelligence - PII extraction, NER, classification, sentiment analysis
Build autonomous AI agents that reason, plan, and execute complex workflows within your applications.
- Agent Framework - Core classes (
Agent,AgentBuilder,AgentExecutor,AgentRegistry) with identity, capabilities, and execution options - Multi-Agent Orchestration - Pipeline, Parallel, Router, and Supervisor orchestrators for composing agent workflows
- Planning Strategies - ReAct, Chain of Thought, Tree of Thought, Plan and Execute, and Reflection handlers
- 56 Built-in Tools - Production-ready tools for data, text, numeric, security, utility, I/O, and network operations
- 18 Agent Templates - Pre-configured specialists for common agent patterns
- Agent Skills Protocol - Reusable task specialists with SKILL.md specification and semantic matching
- Function Calling - Let models dynamically invoke your application's methods with structured parameters
- Tool Registry - Define and manage collections of tools agents can use
- MCP Client Support - Full Model Context Protocol implementation with HTTP/SSE and Stdio transports
- Agent Memory - Persistent memory that survives across conversation sessions
- Reasoning Control - Adjust reasoning depth for models that support extended thinking
- Resilience Policies - Retry, CircuitBreaker, Timeout, RateLimit, Bulkhead, Fallback, and Composite policies
- Agent Delegation - Agent-to-agent delegation with
DelegationManagerand routing
Process and understand content across text, images, documents, and audio.
- Vision Language Models (VLM) - Analyze images, extract information, answer questions about visual content
- VLM-Based OCR - High-accuracy text extraction from images and scanned content
- Speech-to-Text - Transcribe audio with voice activity detection and multi-language support
- Document Processing - Native support for PDF, DOCX, XLSX, PPTX, HTML, and image formats
- Image Embeddings - Generate semantic representations of images for similarity search
- Image Segmentation - Isolate subjects from backgrounds and segment image regions
Ground AI responses in your organization's knowledge with a flexible, extensible RAG framework.
- Modular RAG Architecture - Use built-in pipelines or implement custom retrieval strategies
- Built-in Vector Database - Store and search embeddings without external dependencies
- PDF Chat and Document RAG - Chat and retrieve over documents with dedicated workflows
- Multimodal RAG - Retrieve relevant content from both text and images
- Advanced Chunking - Markdown-aware, semantic, and layout-based chunking strategies
- Reranking - Improve retrieval precision with semantic reranking
- External Vector Store Integration - Connect to Qdrant and other vector databases
Transform unstructured content into structured, actionable data.
- Schema-Based Extraction - Define extraction targets using JSON schemas or custom elements
- Named Entity Recognition (NER) - Extract people, organizations, locations, and custom entity types
- PII Detection - Identify and classify personal identifiers for privacy compliance
- Multimodal Extraction - Extract structured data from images and documents
- Layout-Aware Processing - Detect paragraphs and lines, support region-based workflows
- Schema Discovery - Automatically generate extraction schemas from sample documents
Analyze and understand text and visual content.
- Sentiment and Emotion Analysis - Detect emotional tone from text and images
- Custom Classification - Categorize text and images into your defined classes
- Keyword Extraction - Identify key terms and phrases
- Language Detection - Identify languages from text, images, or audio
- Summarization - Condense long content with configurable strategies
- Sarcasm Detection - Recognize ironic or sarcastic nuances
Generate and refine content with precise control.
- Conversational AI - Build context-aware chatbots with multi-turn memory
- Constrained Generation - Guide model outputs using JSON schemas, templates, or custom grammar rules
- Translation - Convert text between languages with confidence scoring
- Text Enhancement - Improve clarity, fix grammar, adapt tone
Tailor models to your specific domain.
- Fine-Tuning - Train models on your data with LoRA support
- Dynamic LoRA Loading - Switch adapters at runtime without reloading base models
- Quantization - Optimize models for your deployment constraints
- Training Dataset Tools - Prepare and export datasets in standard formats (ShareGPT, etc.)
LM-Kit.NET ships with domain-tuned models optimized for real-world tasks and maintains broad compatibility with models from leading providers. New model architectures are added continuously as they become available in the open-source community.
| Category | Models |
|---|---|
| Text Models | LLaMA, Mistral, Mixtral, Qwen, Phi, Gemma, Granite, DeepSeek, Falcon, GPT-OSS, SmolLM, and more |
| Vision Models | Qwen-VL, MiniCPM-V, Pixtral, Gemma Vision, LightOnOCR |
| Embedding Models | BGE, Nomic, Qwen Embedding, Gemma Embedding |
| Speech Models | Whisper (all sizes), with voice activity detection |
Browse production-ready models in the Model Catalog, or load models directly from any Hugging Face repository.
LM-Kit.NET automatically leverages the best available acceleration on any hardware. Inference performance is continuously optimized with each release through kernel improvements, memory management enhancements, and backend updates.
- NVIDIA GPUs - CUDA backends with optimized kernels (CUDA 12 and 13)
- AMD/Intel GPUs - Vulkan backend for cross-vendor GPU support
- Apple Silicon - Metal acceleration for M-series chips
- Multi-GPU - Distribute models across multiple GPUs
- Hybrid Inference - CPU+GPU inference for models exceeding VRAM capacity
- CPU Fallback - Optimized CPU inference with AVX/AVX2 support
Choose the optimal inference engine for your use case:
- llama.cpp Backend - Broad model compatibility, memory efficiency
- ONNX Runtime - Optimized inference for supported model formats
- OpenTelemetry Integration - GenAI semantic conventions for distributed tracing and metrics
- Inference Metrics - Token counts, processing rates, generation speeds, context utilization, perplexity scores
- Agent Telemetry - Agent name, ID, description, tool invocation events with tool name and call ID
- Event Callbacks - Fine-grained hooks for token sampling, tool invocations, and generation lifecycle
| Platform | Requirements |
|---|---|
| Windows | Windows 7 through Windows 11 |
| macOS | macOS 11+ (Intel and Apple Silicon) |
| Linux | glibc 2.27+ (x64 and ARM64) |
Compatible from .NET Framework 4.6.2 through .NET 10, with optimized binaries for each version.
LM-Kit.NET ships as a single NuGet package with absolutely no external dependencies:
dotnet add package LM-Kit.NETNo Python runtime. No containers. No external services. No native libraries to manage separately. The entire AI stack runs in-process within your .NET application.
- Semantic Kernel - Use LM-Kit.NET as a backend for Microsoft Semantic Kernel
- Vector Databases - Integrate with Qdrant via open-source connectors
- MCP Servers - Connect to Model Context Protocol servers for extended tool access (HTTP/SSE and Stdio transports)
using LMKit;
using LMKit.Model;
// Load a model
var model = new LM("path/to/model.gguf");
// Create a conversation
var conversation = new MultiTurnConversation(model);
// Chat
var response = await conversation.SubmitAsync("Explain quantum computing briefly.");
Console.WriteLine(response);using LMKit.Model;
using LMKit.Agents;
using LMKit.Agents.Tools.BuiltIn;
using LMKit.Agents.Templates;
var model = LM.LoadFromModelID("gptoss:20b");
// Connect to an MCP server
var mcpClient = await McpClient.ForStdio(new StdioTransportOptions
{
Command = "npx",
Arguments = ["-y", "@modelcontextprotocol/server-github"]
});
var agent = AgentTemplates.Assistant()
.WithModel(model)
.WithTools(BuiltInTools.All())
.WithTools(mcpClient)
.Build();
var result = await agent.ExecuteAsync(
"Extract all vendor names and amounts from the attached invoices, " +
"then calculate the total payable this month."
);Running inference locally provides inherent security advantages:
- No data transmission - Content never leaves your network
- No third-party access - No external services process your data
- Audit-friendly - Complete visibility into AI operations
- Air-gapped deployment - Works in disconnected environments
This architecture simplifies compliance with GDPR, HIPAA, SOC 2, and other regulatory frameworks.
© LM-Kit - All rights reserved.