Agent Development Kit
for .NET

An open-source, code-first .NET framework for building, evaluating, and deploying sophisticated AI agents with flexibility and control.

Get Started GitHub 
using GoogleAdk.Core;
using GoogleAdk.Core.Agents;
using GoogleAdk.Core.Runner;

AdkEnv.Load();

var agent = new LlmAgent(new LlmAgentConfig
{
    Name = "quickstart",
    Model = "gemini-2.5-flash",
    Instruction = "You are a concise, helpful assistant."
});

await ConsoleRunner.RunAsync(agent);

  • Agents

    Core execution logic powered by LLMs like LlmAgent, or structural orchestrators like SequentialAgent and ParallelAgent.

    Models & Agents

  • Tools & Plugins

    Actions an agent can take (e.g., Search, API calls) and middleware (e.g., metrics, logging) generated right from your C# code.

    Tools Overview

  • Runners

    Execution engines handling state, streaming, and context. Run in a console, web UI, or background service.

    Running Agents

  • Agent-to-Agent (A2A)

    Built-in support for Remote Agent-to-Agent communication using the A2A protocol (Client and Server).

    Agent-to-Agent

  • Model Context Protocol (MCP)

    Full Model Context Protocol integration allowing dynamic tool discovery and external system connections.

    MCP

  • Evaluations

    Test your agents effectively with Rubric-based, Hallucination, and LLM-as-a-Judge evaluators out of the box.

    Evaluation

  • Orchestration

    Coordinate multiple agents to solve complex workflows using Sequential, Parallel, and Loop architectures.

    Orchestration

  • Memory & State

    Give agents persistent context, session history management, and the ability to recall previous interactions.

    Memory

  • Planning

    Equip your agents with goal-oriented planning capabilities to break down tasks and execute them step-by-step.

    Planning

  • Caching

    Prompt and context caching using Gemini-backed implicit caching to lower latency and costs.

    Caching

  • Plugins & Telemetry

    Lifecycle hooks, streaming events, and OpenTelemetry-style tracing for deep observability.

    Plugins

  • Skills

    File-based modular skills for agents using the SkillLoader and standard SKILL.md configurations.

    Skills

  • Text-to-Speech (TTS)

    Generate audio directly from LLMs like gemini-2.5-flash-preview-tts alongside your responses.

    TTS

  • Code Execution

    Allow your agents to securely execute generated code and analyze the outputs dynamically in real-time.

    Code Execution