generativeai / agents / README.md

Agents framework

An Agent is a system that leverages an AI model to interact with its environment in order to achieve a user-defined objective. It combines reasoning, planning, and the execution of actions (often via external tools) to fulfill tasks.

Resources:

“Agentic" solutions

ADK+Agent Engine (GCP)

Strands+AgentCore (AWS)

Kagent (Kubernetes/Openshift)

Databricks (AgentBricks)

Coming soon

GenAI Frameworks

Langchain (modular)

Build custom LLM agents using reusable components. Design flexible, logic-drivenagent flows.

  • Tool chaining
  • Memomy modules
  • Agent execution

CrewAI (Collaborative)

Mutli-agent system with role assigment and task coordination. Ideal for building agent teams with structure.

  • Task orchestration
  • Role distribution
  • Agent teamwork

AutoGen (Microsoft)

Enable LLM-to-LLM and user-LLM collaboration via dialogue. Great for multi-turn LLM planning tasks.

  • Assistant-user loops
  • Structured dialogue planning
  • Tool support

MetaGPT (Engineering)

Simulates dev teams to build structured software with agents.

  • Roles for Pm, dev, QA
  • Design-first approach
  • Output validation

LangGraph (reactive)

Graph-based execution model for reactive, stateful flows. Excellent for memory and loop heavy logic.

  • Node-based task flow
  • Cycles and retries
  • Multi-agent workflows

AgentOps (Monitoring)

Track and analyze agent behavior in production. Real-time dasboards for running agents.

  • Agent health metrics
  • Logging and debugging
  • Performance alerts

Superagent (open-source)

Drop-in platform with built-in-tools UI and API endpoints. Fast ssndbox for agent experiments.

  • VectorDB + memory
  • REST API access
  • UI for agent interaction

Haystack agents (dev-centric)

Optimized for RAG pipelines and reasoning agents. Best suited for search + logic-based agents.

  • Modular piplines
  • LLML integration
  • Multi-turn task

References

Learning