Agent Frameworks are developer tools, SDKs, and libraries for building AI agents programmatically. These frameworks provide the foundational infrastructure, APIs, and development tools needed to create, customize, and deploy AI agents. They are designed for developers and technical teams who want to build agents with code, offering flexibility, extensibility, and deep integration capabilities.
Revolutionary platform for creating and managing autonomous AI agent teams that collaborate on complex tasks. Features sophisticated agent communication protocols and task management systems. Enables creation of specialized AI agents with defined roles and responsibilities. Includes advanced workflow automation and task delegation capabilities. Supports various use cases from customer service to data analysis. Features monitoring and optimization tools for agent performance. Includes integration capabilities with existing business systems and workflows. Particularly valuable for organizations seeking to automate complex, multi-step processes. Offers detailed analytics and reporting on agent activities and outcomes.
Pricing: Free trial available. Pro plans from $49/month. Enterprise licensing with custom features.
Leading framework for building AI agents and LLM-powered applications. Features agent orchestration, memory management, and chain composition tools. Includes extensive integrations, autonomous agent capabilities, and developer tools. Powers thousands of production AI applications.
Pricing: Open source core. Enterprise: Contact for pricing.
Open-source framework for building agents and workflows over private data, with indexing, retrieval, and event-driven agent workflows. The company's commercial focus is LlamaParse, agentic document processing for parsing, extraction, and indexing of complex files, used in finance, insurance, and healthcare pipelines.
Pricing: Open-source core with paid enterprise solutions
Open-source framework for building and orchestrating AI agents that handle complex tasks reliably. Features graph-based architecture with nodes representing components and edges representing information flow. Includes support for multi-agent systems, human-in-the-loop integration, breakpoints for debugging, time travel functionality, persistence layer through checkpointers, memory systems, and streaming. Offers LangGraph Platform (formerly LangSmith Deployment) for deploying and managing long-running agent workloads at scale. Used by major companies including Rakuten, GitLab, Elastic, Klarna, Cisco, and The Home Depot. Particularly valuable for developers building production agent implementations with fine-grained control over behavior and decision-making.
Pricing: Open-source framework (free). LangGraph Platform: Pricing available upon request for enterprise deployment and management features.
Type-safe Python framework for building production AI agents and LLM applications. Model-agnostic (OpenAI, Anthropic, Gemini, Bedrock, Vertex, etc.). Features MCP, agent-to-agent communication, streaming structured outputs, graph-based workflows, human-in-the-loop tool approval, and durable execution. Integrates with Pydantic Logfire for observability and evals. Open source.
Framework for building, orchestrating, and deploying AI agents and multi-agent workflows. Supports Python and .NET. Complements AutoGen with focus on production orchestration and deployment.
Open-source Microsoft framework for building AI agents with orchestration, memory, plugins/tools, and multi-agent patterns. Designed for production-grade agent applications across Python, .NET, and Java with enterprise-friendly integration points.
Pricing: Open source; enterprise support via ecosystem; see site.
Open-source Python SDK for building multi-agent systems with memory, knowledge, tools, and team workflows, paired with AgentOS: a self-hosted runtime and control plane that runs in your cloud with sessions, tracing, evaluations, approvals, scheduling, and learning for customer-facing and internal agents.
Pricing: Open source and platform offerings; see site.
TypeScript-focused agent framework for building AI workflows with tools, memory, and orchestration primitives. Targets production web stacks that need structured, testable agent behavior beyond one-off prompt chains.
Pricing: Open source and hosted options; see site.
Open-source framework and platform for stateful agents with long-term memory, from the MemGPT team. Agents persist context across sessions, manage memory programmatically (including git-based context repositories), and learn from experience; Letta also runs as a research lab working on memory models for continual learning.
Pricing: Open source and platform options; see site.
Open-source browser automation framework that combines code-first control with AI-powered actions for dynamic web workflows. Provides primitives for acting on pages, extracting structured data, and running multi-step browser tasks with better resilience than brittle selector-only scripts. A strong fit for teams building agentic browser workflows in production codebases.
Open-source library for building AI agents that can navigate and interact with websites to complete multi-step tasks. Supports browser-driven automation patterns such as form completion, information retrieval, and workflow execution with LLM-guided decision loops. Useful for developers building autonomous web-task agents with Python-first tooling.
Pricing: Open-source self-hosted option available. Cloud plans available; see site for current pricing.
Open-source research community and Python stack for multi-agent systems, workforce orchestration, and agent RL. Ships libraries such as CAMEL, OWL, OASIS, and environments for simulation, data generation, and real-world task automation at scale.
Pricing: Open source; hosted and enterprise offerings via community and partners; see site.
Lightweight open-source SDK for building multi-agent AI applications in Python and TypeScript. Features agent loop with tool invocation, handoffs between specialized agents, guardrails, function tools with Pydantic validation, sessions, tracing, voice agents, and sandbox agents for isolated execution. Provider-agnostic across OpenAI Responses, Chat Completions, and 100+ LLMs. Production successor to Swarm; 27k+ GitHub stars.
Google's open-source, code-first framework for building, evaluating, and deploying AI agents and multi-agent workflows. Features workflow runtime with sequential, parallel, and loop orchestration; Task API for agent-to-agent delegation; rich tool ecosystem with MCP; and SDKs in Python, TypeScript, Go, and Java. Model-agnostic and deployment-agnostic-optimized for Gemini and Google Cloud but compatible with other models and runtimes. 20k+ GitHub stars on adk-python.
Open-source Hugging Face Python library for building agentic systems with minimal code. CodeAgent writes actions as executable Python-loops, conditionals, and tool calls-rather than JSON tool payloads; sandbox execution via Modal, E2B, or Docker. Model-agnostic across Hub inference, OpenAI, Anthropic, LiteLLM, Transformers, and Ollama. Hub integration for sharing agents and tools; CLI utilities for quick runs. Successor to transformers.agents; 27k+ GitHub stars.
Model-agnostic Python SDK for building production AI agents with tool use, MCP, multi-agent workflows, and OpenTelemetry observability. Supports Bedrock, Anthropic, OpenAI, Gemini, Ollama, and LiteLLM providers. AWS-originated; runs on any cloud or on-prem.
Open-source multi-agent framework from DeepWisdom that simulates a software company: product, architect, engineer, and QA agents follow SOPs to turn a one-line requirement into code, repos, and docs. The team has since productized it as MGX and now Atoms, a commercial vibe-building business team, while MetaGPT remains open source.
Linux Foundation open-source framework for production multi-agent systems in Python and TypeScript. Constraint-based agents, declarative orchestration, MCP/A2A support, and OpenTelemetry observability under open governance.
Anthropic's official SDK for building Claude-powered agents in Python (with TypeScript counterparts). Tool use, computer control patterns, and agent loops aligned with Claude Code / Claude platform capabilities.
Pricing: Open source. Free. Claude API usage billed separately.
Open-source TypeScript AI agent framework with observability, deployment, and guardrails tooling. Built for engineering teams shipping typed agent apps beyond one-off LLM scripts.
Pricing: Open source. Free. Cloud/platform options-see site.
Vercel's TypeScript toolkit for building AI-powered apps and agents-streaming, tool calling, and multi-step generation across major model providers. Widely used to ship agents inside Next.js and web stacks.
Pricing: Open source. Free. Infrastructure usage via your model providers and Vercel.
Context infrastructure for AI agents inside software products. Builds a product's context once, adapts it privately to each customer's rules, people, and decisions, and keeps it learning so agents stay accurate. Ranked #1 on the Spider 2.0 and LiveSQL text-to-SQL benchmarks.
Open-source multi-agent AgentOS formerly known as AutoGen, maintained by the original AutoGen authors under ag2ai. Actively developed alternative to Microsoft's maintenance-mode AutoGen repo; AG2 Classic preserves ConversableAgent patterns while AG2 1.0 adds a network-based architecture.
Enterprise-grade multi-agent orchestration framework and agent marketplace. Build, compose, and run swarms of specialized agents with production-oriented tooling for Python teams shipping multi-agent systems.
Pricing: Open-source framework; marketplace and cloud options-see swarms.ai.
Open-source multi-agent framework to spin up role-based AI workforces with less boilerplate. Orchestrates agents, tools, and workflows for research, coding, and ops automation; Python-first with docs and examples for rapid crews.
Modular Python framework for AI agents, pipelines, and workflows with secure connections to enterprise data and APIs. Structured tasks, tools, and memory for production agentic systems.
Pricing: Open-source framework; cloud and enterprise options-see griptape.ai.
Open-source framework for building realtime voice, video, and multimodal AI agents on LiveKit. Python and Node SDKs for speech pipelines, tool use, and low-latency conversational agents.
Pricing: Open-source agents SDK; LiveKit Cloud for hosted realtime infrastructure.
Open-source framework from Daily for voice and multimodal AI agents. Composable pipelines for speech-to-speech, transports, and tool-calling agents used widely in realtime voice products.
LangChain's batteries-included agent harness on LangGraph. Ships planning, subagents, filesystem context, and human-in-the-loop middleware so teams can run deep multi-step agents without assembling primitives from scratch.
Open-source framework for putting AI agents inside real applications, and creator of AG-UI, the open agent-user interaction protocol adopted by Google, Microsoft, Amazon, and Oracle. React and Angular SDKs for generative UI, shared state, frontend tool calls, and human-in-the-loop, connecting to LangGraph, CrewAI, Mastra, PydanticAI, and other backends. Libraries see 4M+ weekly downloads; raised a $27M Series A in 2026.
Pricing: Open source. Free. Enterprise and self-hosted tiers; see site.