Agent frameworks & SDKs

Developed by Microsoft, this multi-agent application framework supports autonomous agent collaboration and human-in-the-loop workflows for building next-generation large language model applications.

Python framework for orchestrating role-playing autonomous AI agents, offering Crew and Flow abstractions to build collaborative multi-agent workflows for production automation.

Open-source API layer that turns local models into production-grade AI applications, offering document retrieval, tool calling and model serving for private, leak-free development.

A developer framework and runtime for building agent platforms, with SDK, AgentOS serving, storage, observability, scheduling and management UI.

Frontend development kit adding chat, generative user interfaces, and shared state to AI agents. Compatible with arbitrary agent frameworks and React, Angular, mobile, Slack, and other platforms.

Implements an event-driven, durable workflow engine that provides resilient execution for microservices and AI agents, aimed at engineers orchestrating distributed applications.

Gives TypeScript developers a framework for building production AI agents and applications. It provides model routing, agents, workflows, memory, and MCP integrations for rapid development.

Enables Python developers to build MCP servers, clients, and interactive applications with idiomatic APIs. The framework handles schemas, authentication, and transport details defined by the protocol.

Supplies a vendor-neutral TypeScript toolkit for connecting to major language models. Developers use it to build AI applications, agents, and conversational interfaces with unified APIs.

Offers an open-source Python framework for building production-grade RAG systems. Developers use it to implement semantic search, question answering, and LLM agent pipelines.

The official Python SDK for building MCP servers and clients that connect LLMs to tools.

A Go framework for building agents and services with models, memory, tools, planning, guardrails, and discovery, supporting exposure of agents and services via MCP and A2A protocols.

Open-source Python framework for developers to build, evaluate, and deploy complex AI agents, supporting multi-agent orchestration and integration with external tools and services.

A Python agent framework centered on Pydantic's type system, supporting multi-model switching, tool calling, memory, and collaboration features for building structured LLM-powered agents.

Open-source framework for developers building enterprise customer-service conversational agents, using rule constraints and context engineering to deliver controllable, consistent, traceable LLM interactions.

A JavaScript framework for building LLM applications, providing standardized interfaces for models, embeddings, and vector stores plus chaining and orchestration for complex workflows.

Early multi-agent development framework helping developers build agent applications through role-playing, multi-agent collaboration, and simulation capabilities for research and practical use cases.

Agent development framework built on Qwen large models, providing model abstraction, tool calling, memory components, browser assistant examples, and backend for Qwen Chat.

Open-source durable execution platform for TypeScript developers to build and deploy AI agents and workflows, with support for long-running jobs, retries, scheduling, observability, and autoscaling.

Lightweight developer library offering a simple unified interface to multiple generative AI providers, with conversation abstractions and tool-calling support for building agents.

Open-source framework for developers building next-generation AI chatbots and assistants, offering integrations, SDKs, and CLI building blocks to streamline conversational AI development and deployment.

Unified development framework for enterprise RAG that combines a 300+ small-model catalog with document parsing, chunking, retrieval, and local deployment for knowledge applications.

Open-source framework for developers building real-time voice AI agents, providing multimodal conversations, model integrations, task scheduling, and telephony access capabilities.

Open-source general framework for Python and .NET to build, orchestrate, and deploy production-grade AI agents and multi-agent workflows for developers building enterprise solutions.

Official TypeScript SDK for Model Context Protocol servers and clients, providing transports, authentication, and framework-integration middleware for developers building MCP applications.

Open-source Java library bringing LLM abstractions to the JVM ecosystem. Unifies major models and vector stores while simplifying tool calls, agents, and RAG development.

Go framework for building LLM applications, offering component abstractions, agent development kits, and graph-based workflow orchestration for developers creating AI agents.

All-in-one Python AI framework for vector search, RAG, LLM orchestration, and agentic workflows. Supports multimodal embeddings and semantic search for developers building knowledge applications.

Python framework for building conversational AI applications, offering ready-made chat UI, decorators, and run commands for production-ready deployment, integrating major LLMs and retrieval tools for developers.

An incremental indexing framework keeping code and docs as continuously fresh context for AI agents.

Developer-focused LLM application engineering suite for orchestrating flows of models and tools while supporting debugging, evaluation, and deployment in end-to-end workflows.

An open-source toolkit for building AI SRE agents, offering tool integrations, customizable workflows, and reproducible production-incident benchmarks for evaluating reliability automation.

Minimalist LLM development framework in about a hundred lines of code providing graph abstractions for developers to quickly build agents, workflows, and retrieval-augmented generation applications.

A React SDK and backend for building generative-UI agents that stream and render interactive components.

Open-source framework for building real-time multimodal conversational AI and voice agents, providing examples, RTC extensions, and hardware-software integration for interactive application developers.

Open-source multi-agent development framework that generates agent teams from natural-language goals, featuring evolution loops, MCP integration, and over a hundred tools for AI application builders.

Production-grade agent development framework for Java developers supporting multi-agent orchestration, graph workflows, and model-tool invocation to build enterprise AI applications quickly.

Full-stack MCP development framework for TypeScript that scaffolds, builds, tests, and publishes MCP servers and interactive views consumed by AI agents.

Open-source TypeScript framework for developers building AI agents, providing memory, tool integration, multi-agent orchestration, and observability tooling for production-ready applications.

Multi-platform, multi-language SDK for embedding the GitHub Copilot agent runtime into applications and services through programmatic calls, helping developers add agentic coding assistance to their own products.

Go port of LangChain providing LLM calls, prompt chaining, memory management, and agent components for developers building LLM-powered applications in Go.

High-performance AI pipeline engine for LLM application developers, combining a C++ core with extensible Python nodes to build, debug, and scale workflows across many models and vector databases.

A complete Go implementation of the MCP protocol, providing server and client SDKs to help developers build tool and resource services integrating LLM applications with external data.

Provides a multi-agent development framework for developers with SDKs, supporting over a hundred large language models, memory management, and workflow orchestration to build self-reflective autonomous agents.

Supplies an open-source agent development kit for Go developers, offering multi-agent orchestration, tool calling, evaluation utilities, and cloud-native deployment support for production applications.

Modular Rust library for building extensible LLM applications, providing multi-model interfaces, vector retrieval, and agent orchestration runtimes for developers efficiently.

Offers a development framework for building agents on the Model Context Protocol, featuring composable workflow patterns, tool integration, and persistent execution for reliable multi-step automation.

Supplies a programmable sandboxed framework for building AI agents, with sessions, tools, skills, and deployment options across environments for developers prototyping contained autonomous workflows.

Offers a Python agent development framework for developers to build autonomous and traditional agents, supporting custom tools, MCP integration, and document OCR processing for practical workflows.

Provides a multi-agent routing and orchestration framework for developers, dispatching tasks across different model-powered agents while managing conversational context across cloud and edge runtimes in three languages.

Open-source SDK for Python and TypeScript developers to build and run AI agents with tool calling, multi-agent collaboration, and memory using only a few lines of code.

Python library for LLM-driven multi-agent persona simulation, letting researchers customize virtual characters that interact in simulated worlds for ad evaluation, software testing and brainstorming insights.

TypeScript framework for multi-agent development that lets Node.js developers define agent teams, while a coordinator runtime plans task DAGs and handles scheduling, approvals, tracing, evaluation, and checkpoint recovery.

Open-source self-hosted cloud-native framework for managing collaborative AI agent teams, supporting extensible skills and usable as a Python library or API for developers building multi-agent applications.

An open-source framework for building agent-native applications with business interfaces, providing actions, data access, authentication, and team collaboration primitives for developers.

Defines a filesystem-centered framework for building persistent AI agents, offering directory conventions, scaffolding, and multi-channel scheduling to help developers create long-lived, maintainable agent systems.

Provides the official Go SDK for developers building Model Context Protocol clients and servers, maintained with Google to support agent tool integration workflows.

Provides an open-source programming framework from the AutoGen creators for developers building AI agents and coordinating cooperation among multiple agents to solve tasks.

Offers a Kotlin and JVM framework for enterprise developers building predictable fault-tolerant AI agents, with tool calling, persistence, and deployment across backend, mobile, and browser environments.

Collects thirty-five production-grade agentic AI architectures into a Python library and runnable textbook, with unified multi-provider LLM interfaces and a seventeen-task benchmark leaderboard for developers.

Provides the official C# SDK for .NET developers building Model Context Protocol clients, servers, and hosts, with support for dependency injection and agent application workflows.

JVM agent framework letting developers orchestrate LLM-driven, goal-oriented workflows with code and domain models for enterprise Java and Kotlin applications.

Offers a Ruby framework for developers building chatbots, AI agents, RAG applications, and multimodal workflows, unifying access to major LLM providers through expressive idiomatic code.

Agent development framework for coding, building, and evaluating AI workflows, supporting multiple models, comprehensive MCP features, skills, ACP, A2A protocols, and replay-based evaluation.

Development framework simplifying construction of multi-agent LLM applications, supporting low-cost assembly, rapid iteration, deployment, and fine-tuning for developers building complex workflows.

Real-time transport framework for Java-based AI agents, allowing developers to build once with annotations and deliver interactions over WebSocket, SSE, gRPC, and WebTransport while supporting MCP, A2A, and AG-UI protocols.

Python SDK for developers building AI agents on Google Antigravity and Gemini, exposing tool-calling capabilities and full platform features for agent development.

Gives developers a production-ready framework in Python and TypeScript for building multi-agent workflows with memory management, tool use, and standard agent protocol integrations.

Simplifies building MCP servers in TypeScript by providing helpers for defining tools, resources, and prompts, along with authentication, routing, and streaming transport best practices.

A multi-agent AI framework in .NET for developers building conversational robots and assistants, with support for multiple LLM providers, RAG pipelines, and modular plugin-based agent design.

The official specification and SDK repository for the MCP Apps protocol. It defines how MCP servers deliver embedded UIs such as charts and forms rendered inline in AI chat clients.

A multi-turn reinforcement-learning framework for training LLM reasoning agents in interactive stochastic environments. It includes diagnostic tools and support for multiple evaluation environments.

A reusable Android voice AI SDK built with Kotlin and MVVM architecture that adds a complete voice-driven conversational pipeline, from speech input to LLM responses, to any existing application within minutes.

Enterprise-grade TypeScript framework for building, testing, and deploying production-ready MCP servers and AI-native applications, aimed at teams shipping maintainable backend services.

Self-evolving embodied AI framework for physical and robotic tasks, providing infrastructure for skill execution, verification, and task-level evaluation of agentic workflows.

A node-based LLM framework inside ComfyUI for building workflows and multi-agent collaborations. It connects to MCP servers, speech, OCR, image models, messaging platforms, and OpenAI-compatible LLMs.

Offers Espressif's on-device AI agent framework for ESP32-based IoT products, letting developers define behaviors through chat while combining local sensing with edge decision-making.

Python library for writing predictable generative programs, using type annotations and automatic retries to implement structured large-model call workflows for developers.

Portable protocol components and governance engine for AI reasoning, RAG, and agent workflows, including problem maps and debugging cards. Aimed at improving reliability in real-world deployments.

Open specification defining a minimal standard for packaging AI agent extensions into distributable plugins, covering portable formats and manifests for skills and MCP services.

Provides a lightweight open-source framework for building AI agents with memory, MCP integration, and skills, supporting multi-agent collaboration, self-learning, and major large language models.

Single-agent Python core framework providing a pluggable reasoning loop with tool permissions and observable event streams for building controllable agents.

Decentralized multi-agent framework modeling agent selection as an online bandit problem, allowing specialized roles to emerge through collaborative reasoning without central coordination.

Open-source voice-AI SDK, MIT-licensed in Python and TypeScript, that gives AI agents phone numbers in four lines of code. Provides a self-hosted agent loop with Twilio, Telnyx, and Plivo integrations for builders owning the full telephony stack.

Production-ready framework for building, deploying, and scaling secure MCP servers that power AI agents. Includes built-in authentication, observability, telemetry, debugging tools, and runtime support for teams shipping real-world agent infrastructure.

Development framework for building agents over enterprise OA, ERP, CRM, and ticketing systems, enabling AI to invoke real business capabilities under permission, audit, and governance constraints.

Offers a production-oriented AI workflow runtime for building, validating, recovering, and shipping complex AI workflows as dependable, service-ready deployments with deterministic orchestration.

A scalable reinforcement-learning framework for training diverse AI agents inside sandboxed environments, connecting rollout generation with gradient updates for researchers building agentic reinforcement-learning pipelines.

A Swift-native runtime for building production AI agents with type-safe tools, workflows, memory and guardrails.