AI Models & Technology · AI Agents
What frameworks do developers actually use to build AI agents
Developers building AI agents commonly use frameworks like LangChain and its LangGraph extension for orchestrating multi-step workflows, CrewAI for coordinating multiple specialized agents, and provider-specific SDKs, increasingly connected via the shared Model Context Protocol standard.
Key takeaways
- LangChain and its LangGraph extension are widely used for orchestrating multi-step agent workflows, including ones that need to pause and resume.
- CrewAI is built specifically around coordinating multiple specialized agents working together on a shared task.
- The Model Context Protocol (MCP) has become a shared standard for connecting agents to external tools, letting a tool built once work across multiple frameworks.
- Provider-specific SDKs from major AI companies offer another path, tightly integrated with that provider's own models.
Why Frameworks Exist at All
Building an AI agent from scratch means handling a lot of repeated infrastructure — managing conversation state, calling external tools, handling errors mid-task — so most developers build on top of an existing framework rather than writing all of that plumbing themselves.
Workflow Orchestration: LangChain and LangGraph
LangChain, and its LangGraph extension in particular, is widely used for orchestrating multi-step agent workflows — LangGraph specifically supports things like pausing a workflow partway through for human review and then resuming it, which matters for agents handling longer-running or higher-stakes tasks.
Coordinating Multiple Agents: CrewAI
CrewAI is built specifically around scenarios with multiple specialized agents working together on a shared task, providing structure for defining each agent’s role and how they hand off work to each other — a more specialized fit than a general single-agent framework for genuinely multi-agent projects.
A Shared Standard for Connecting Tools: MCP
The Model Context Protocol has become a widely adopted shared standard for connecting an agent to external tools and data sources — a tool built to the MCP standard once can be used across multiple different frameworks, reducing the need to rebuild the same tool integration separately for each one.
Provider-Specific SDKs as a Simpler Starting Path
Beyond the general-purpose frameworks, SDKs offered directly by AI providers give developers a simpler, more tightly integrated starting path when building an agent around one specific provider’s models, trading some framework flexibility for a lower setup barrier and tighter integration with that provider’s own tools and features.
Bottom Line
Developers building AI agents typically choose between general orchestration frameworks like LangChain/LangGraph, multi-agent-focused frameworks like CrewAI, or provider-specific SDKs — increasingly connected to external tools through the shared MCP standard rather than framework-specific integrations.
Look Up AI Terms
Search plain-English definitions of AI and machine learning terms in our free AI Glossary.
Go deeper
Related questions
- Can AI Agents Work Together Even If They're Built on Different Models?
- What Is a Multi-Agent System, and Why Use Multiple Agents Instead of One?
- What Is the Model Context Protocol (MCP)?
- Can AI Agents Take Actions on Your Behalf, Like Booking Flights?
- How Do AI Agents Decide Which Tool to Use for a Given Task?
- What Happens When an AI Agent Gets Stuck or Fails Mid-Task?
Sources
- [1]Model Context Protocol — Anthropic
- [2]AI agent frameworks — LangChain
Written by Editorial Team
Last updated August 7, 2026
Get one well-sourced answer a week
No spam. Unsubscribe anytime.