Agent Frameworks Directory

FAQ

• What is an agent framework?

  An agent framework is a platform for building, orchestrating, and deploying AI agents and multi-agent workflows. It typically provides orchestration (e.g. graph-based or sequential), tool use and integrations, memory and state, and support for one or more LLMs so you can ship agents without writing everything from scratch.

• What is the difference between an agent and a workflow?

  An agent is LLM-driven — its steps are decided at runtime from context and tools. A workflow is a predefined sequence of steps (often including agents). Use an agent when tasks are open-ended or conversational; use a workflow when the process is well-defined and you need explicit control, governance, or human-in-the-loop. Many systems combine both.

• What is multi-agent RAG? How is it different from single-agent RAG?

  Multi-agent RAG uses several specialized agents (e.g. planner, retriever, reasoner) with a coordinator. Single-agent RAG uses one model for retrieval and generation. Use multi-agent RAG for complex queries, many data sources, or enterprise scale; use single-agent RAG for simple queries and when cost and latency matter most.

• What are the main parts of an agent framework?

  Typical parts include: orchestration (how steps and agents are coordinated), tool use and connectors (APIs, code, search), memory and state (conversation, episodic, long-term), the reasoning engine (LLMs), and often deployment and observability. Frameworks differ in how much they give you out of the box.

• How do I choose an agent framework?

  Consider your use case (orchestration, multi-agent, RAG, or cloud), preferred language and stack, open source vs managed, and how much control you need. This directory groups frameworks by category and links to docs so you can compare and try the ones that fit.

• What's the difference between LangGraph, CrewAI, and AutoGen?

  LangGraph uses graph-based state machines and explicit control flow; it suits complex production workflows and strong governance. CrewAI uses role-based teams and sequential tasks; it's good for rapid prototyping and team-style collaboration. AutoGen (Microsoft) is conversational and event-driven; it fits chat-style and human-in-the-loop use cases.

• When should I use a vertical vs horizontal framework?

  Vertical frameworks are built for specific domains (e.g. finance, healthcare) and offer faster deployment and built-in compliance. Horizontal frameworks are general-purpose and flexible but need more customization. Choose vertical when the domain fit is strong; choose horizontal when you need maximum flexibility across use cases.

• Which framework is best for beginners?

  CrewAI is often cited as the gentlest start: role-based agents and clear docs let you build a first multi-agent flow quickly. LangGraph is more flexible and production-oriented but has a steeper learning curve (graphs and state). Starting with CrewAI for quick wins, then trying LangGraph when you need finer control, is a common path.

• Which framework is best for production or enterprise?

  LangGraph is widely used in production for complex workflows, state management, and observability (e.g. LangSmith). CrewAI can get you to an MVP fast but is often used for internal or prototype use. Choose based on your need for audit trails, compliance, and long-term maintainability.

• When do I need multi-agent instead of a single agent?

  Consider multi-agent when you need clearly different capabilities (e.g. search vs code execution), parallel work, isolation (e.g. different credentials or data access), or reusable specialist agents. Prefer a single agent when the task is cohesive, latency matters, or you want simpler state and debugging.

• Which framework fits orchestration, multi-agent, RAG, or cloud use cases?

  This directory tags frameworks by focus: orchestration (workflows and state machines), multi-agent (teams and roles), RAG (retrieval-augmented generation), and cloud (managed platforms). Use the category filters at the top of this page to match your use case.

• Which LLM providers do agent frameworks support?

  Most frameworks support several providers: OpenAI, Azure OpenAI, Anthropic (Claude), Google (Gemini/Vertex), Amazon Bedrock, and often Ollama or other local/OpenAI-compatible APIs. Check each framework's docs for the exact list and how to configure and switch providers.

• How do frameworks handle memory and state?

  Frameworks typically provide session or conversation state, sometimes episodic or long-term memory, and context providers. Implementations vary: some use vector stores and hybrid search, others use built-in state in graphs or workflows. See each framework's documentation for details.

• What about tool use and integrations (APIs, databases, MCP)?

  Frameworks usually offer tool-calling APIs, schema generation, and integrations (APIs, databases, MCP, etc.). Capabilities differ: some support sandboxing and approval flows, others focus on connectors. Check the framework's docs and the official site links in this directory.

• How do I get started with a specific framework?

  Each listing links to the framework's official site. From there, navigate to the docs, quickstarts, and API references — most frameworks have a 'Get started' section.

• How can I suggest or submit a framework to be listed?

  Use the submit page to suggest a framework. We review submissions and add frameworks that fit the directory.

• How are frameworks categorized (orchestration, multi-agent, RAG, cloud)?

  Frameworks are grouped by focus: orchestration (workflows and state machines), multi-agent (teams and roles), RAG (retrieval-augmented generation), and cloud platforms (managed agent services). You can use these categories to find options that match your use case.