Agent SDKs & Frameworks
General agent-building frameworks, SDKs, and toolkits. These ship no coding agent of their own — developers construct agents from them (AutoGen, CrewAI, LangGraph, and similar).
Pioneered multi-agent orchestration patterns at Microsoft Research with a layered extensible architecture (Core -> AgentChat -> Extensions), cross-language...
Combines two complementary paradigms -- autonomous role-based Crews and event-driven deterministic Flows -- letting developers balance agent autonomy with...
A model-centric (not framework-centric) agent service designed for increasingly agentic LLMs, shipping a full FastAPI backend + Web UI with multi-tenancy,...
It is a TypeScript-native agent framework from the Gatsby creators: graph-based workflows with durable human-in-the-loop suspend/resume, Observational Memory, a...
Framework for building LLM-powered applications that helps developers chain together interoperable components and third-party integrations, simplifying AI...
Framework for building LLM applications leveraging Qwen's instruction following, tool usage, planning, and memory capabilities, featuring MCP support, code...
AI agent framework for autonomous, self-improving multi-agent systems. Unique five-layer agent stack (Prompt -> Context -> Harness -> Loop -> Graph) plus a...
Programmable TypeScript harness framework for building autonomous agents where the agent IS a function composed of hooks (useModel, useSandbox, useSkill,...
Official Anthropic Python SDK for building agentic applications powered by Claude Code. In-process SDK MCP servers let custom tools run in the same Python...
Enterprise-grade multi-agent orchestration framework with 60+ prebuilt multi-agent architectures (sequential, concurrent, hierarchical). AutoSwarmBuilder...
Model-driven SDK taking a minimal-code approach to building agents. Model-agnostic (any model, any cloud; swap backends without code changes). Agent loop traces...
'Describe the goal, not the graph' — the coordinator agent builds the task DAG at runtime from the goal rather than hand-wiring workflow graphs. Controlled...
A data-centric, self-evolving autonomous language agent framework (Agents 2.0) that applies the connectionist learning procedure to agent training. Makes an...
The runtime layer that turns an LLM into a working agent — model calls, MCP tools, skills, sandboxing, approvals, context management, and session state —...
Intuitive, lightweight, extensible Python framework (from CMU/UW-Madison researchers) for building LLM-powered multi-agent applications. Agent-as-actor paradigm...
Real-time engine for AI agents on the JVM with deep-agent defaults (memory, plan, virtual filesystem, sub-agent delegation) and governance/compliance (OWASP...
Official Anthropic SDK that programmatically embeds Claude Code's agentic capabilities (codebase understanding, file editing, shell command execution, complex...
OS-level cloud-native runtime and SDK for production AI agents ('The Linux of AI Agents'). Unified Resource Abstraction — same API works across all providers...
Platform for designing and deploying autonomous and semi-autonomous LLM-powered agents and workflows with a no-code visual agent editor for connecting nodes...
A unified Go library for building AI agents with a single API across many LLM providers and models, compiling to native machine code; built to power Charm's...
PyTorch-like runtime for dynamic multi-agent and multi-session workflows — Session as the central flowing value (like a Tensor); composable agents like...
Multi-agent framework driven by a Knowledge Graph (Eventic Knowledge Graph / EKG) engine rather than purely model-based or fixed-flow approaches. Successful...
Agent framework (AutoGPT/LangChain-style) that takes a directive and an array of APIs, then converses with AI until the directive is complete. Universal API...
Python framework for interacting with LLMs — chat, function calling, structured output, RAG, and agentic chains with tools. Key differentiator is guided...
First Langchain Agent creator designed to build, prototype, and deploy AI-powered agents with on-the-fly agent/tool creation, interactive CLI, session...
An embedded agent SDK for B2B SaaS that lets your end customers build their own features and micro-apps on top of your product: agents execute through the host...
Open source AI agent studio with visual drag-and-drop builder, multi-model composition (GPT/Claude/Gemini), GitHub AI Operations automation, and knowledge store...
Unified programming interface for deploying AG2 (AutoGen) multi-agent workflows to production; seamless OpenAPI integration with few lines of code; Tester class...
"Reverse-engineered Rust port of Claude Code architecture as an embeddable SDK; graph memory (Grafeo) with 98us recall vs Claude Code's 7.5s LM call; 6MB...
Python SDK for the Steamship platform to build, scale, and monitor AI Agents, with a triad of deployable components (Agents, Packages, Plugins)....
AI-Native, Intent-Oriented Programming Language built on Python; first framework to formally represent human intent as structured elements (Intent IR: Goal,...
Sagentic.ai Agent Framework, a unified platform for building, running, and scaling autonomous agents. Features a dev server with hot reloading and spawning...
Modular, type-safe agent runtime for building LLM-powered applications ('LangChain + LangGraph for OCaml'). Written in OCaml 5.4 using Eio structured...
DSL and toolchain for authoring AI pipeline workflows with first-class syntax for prompts, shell scripts, model config, branching logic, and node types (agent,...
Production-ready Rust implementation of the 2389 Agent Protocol — a standard for interoperable AI agents communicating via MQTT in distributed systems. Provides...
Was described as an AI Agent Automation Platform for rapidly prototyping, testing, and deploying Multi-Agent Systems from the browser. GitHub repo...
Was described as a framework for building autonomous decision-making agents integrating reinforcement learning, NLP, and multi-agent systems. GitHub repo...
Experimental agent API in Hugging Face Transformers offering CodeAgent (one-shot code generation and execution) and ReactAgent (step-by-step ReAct with JSON or...
Batteries-included agent harness built on LangGraph. Provides planning, file system access, context management, sub-agents with isolated context windows, shell...
Low-level orchestration framework and runtime for long-running, stateful agents; combines deterministic and agentic steps in one graph; durable execution,...
The direct successor to both Semantic Kernel and AutoGen (created by the same teams), merging AutoGen's simple agent abstractions with Semantic Kernel's...
A generalist multi-agent system with a modular, plug-and-play design where agents can be added/removed without reworking the system. An Orchestrator leads four...
Agent SDK / agentic infrastructure library for Rust providing tool execution with structured input/output, MCP client integration, permission-gated approval...
Open-source library for connecting and optimizing teams of AI agents across any framework. Supports MCP as both client and server (FastMCP), a public Plugin...
Tailored for IBM Granite and Llama 3.x models; offers a sandboxed code interpreter, OpenAI-compatible Assistants API, serialization for pausing/resuming complex...
Despite the repo name 'python-agents', this is a TypeScript agent framework (Clarisse Agent Framework) optimized for IBM Granite and Llama 3.x models; includes...
A lightweight, minimal ReAct (Reasoning & Acting) AI agent framework described as 'like Eliza, but tinier, cleaner, and with way less clutter.' Emphasizes...