Agent Libraries and Tools

Content updated May 2026. The agentic framework landscape consolidated significantly through 2025. This page covers the six production-grade options and supporting development tools.

The Six Production Frameworks

By 2025, the multi-agent framework landscape settled around six options. Each reflects a fundamentally different philosophy on how agents should coordinate. Choosing the wrong one for your coordination model is expensive to fix.

LangGraph

Paradigm: Graph/stateful — explicit control flow with checkpointing

LangGraph (from the LangChain team) models agent workflows as directed graphs with explicit state management. Each node in the graph is a function; edges define transitions. Supports cycles (agent loops), checkpointing for long-running tasks, and streaming.

Best for: Complex multi-step workflows with dynamic decision branches; systems where you need fine-grained control over state and need to audit intermediate states.

LangGraph documentation
LangSmith for tracing and evaluation

CrewAI

Paradigm: Role-based — agents with defined personas collaborate

CrewAI lets you define agents with specific roles, goals, and backstories. Agents work together as a "crew" toward a shared objective. The framework handles delegation and communication.

Best for: Workflows that naturally map to a team of specialists (researcher, analyst, writer, reviewer).

CrewAI documentation

OpenAI Agents SDK (March 2025)

Paradigm: Handoff-based — agents pass tasks between themselves

Released March 2025, replacing the experimental Swarm framework. Three built-in primitives:

Handoffs — agent-to-agent task transfer with full context
Guardrails — input and output validation at boundaries
Tracing — end-to-end observability across agent chains

Python-first. Best for: Sequential pipelines where specialised agents handle distinct phases.

Source

OpenAI Agents SDK release, March 2025

OpenAI Agents SDK documentation

Google Agent Development Kit (ADK)

Paradigm: Hierarchical tree — root agent delegates to sub-agents

Google released ADK alongside the A2A protocol in April 2025. A root agent decomposes goals and delegates to sub-agents; the hierarchy can be as deep as needed. Native integration with the A2A protocol for cross-vendor agent communication.

Best for: Enterprise orchestration scenarios with clear task hierarchies; systems that need to interoperate with other vendors' agents via A2A.

Source

Google ADK documentation

AutoGen (Microsoft)

Paradigm: Conversational — agents coordinate through structured dialogue

AutoGen models agents as conversational actors that discuss and coordinate through structured message exchanges. Supports both single-agent and multi-agent scenarios with flexible conversation patterns.

Best for: Exploratory problem-solving where the solution path is not predetermined; research and analysis workflows.

AutoGen documentation

Anthropic Agent SDK

Paradigm: Pipeline-native — built around Claude models and MCP

Anthropic's Agent SDK is designed for building multi-agent pipelines natively on Claude models. Tight integration with MCP (Model Context Protocol) for tool access. Pairs with Claude Computer Use for desktop automation.

Best for: Claude-centric deployments; systems that need native MCP tool integration.

Supporting Tools

agent-service-toolkit

A production-ready template for deploying LangGraph agents as a service, with FastAPI, streaming, and auth baked in.

agent-service-toolkit on GitHub

OmniParser v2 (Microsoft)

Turns any LLM into a computer-use agent by parsing screen content into structured, interactable elements. Enables computer-use without requiring a natively trained computer-use model.

OmniParser v2

Blog — Microsoft Research, 2025.

Framework Selection Guide

Framework is not model

All six frameworks can work with multiple LLM providers. Your framework choice is an architectural decision about coordination patterns, not a model commitment.