AI Agent Frameworks Packages
Frameworks for building LLM agents with tools, memory, and orchestration for multi-step, autonomous workflows (LangChain, LlamaIndex, CrewAI, LangGraph).
Packages in AI Agent Frameworks
Kind
Sort:Used by Most Apps
GitHub Copilot SDK (Rust)
Rust SDK for embedding GitHub Copilot's agentic workflows into your application over JSON-RPC.
baml
The Python runtime for BAML — a typed prompt language that compiles LLM calls into type-safe, testable client code.
Rig
A Rust framework for building portable, modular LLM-powered applications and agents.
pydantic-monty-runtime
The compiled `monty` CLI and worker binary that gives pydantic-monty and @pydantic/monty crash-isolated Python sandbox execution.
Hatchet Python SDK
Official Python SDK for Hatchet, a distributed, fault-tolerant task queue and durable workflow engine.
GEPA
Optimize prompts, code, and agents with LLM reflection and evolutionary search
AgentOps
Observability and monitoring SDK for building, evaluating, and deploying AI agents
Deep Agents (JS)
The batteries-included TypeScript harness for building controllable, production-ready LangGraph agents.
BrowserGym Core
Gymnasium environment for building and evaluating LLM-driven web agents in a real Chromium browser.
LangChain Rust
Build LLM-powered applications in Rust with composable chains, agents, tools, and vector stores.
Strands Agents Tools
A batteries-included tool library for Strands Agents, giving LLM agents file, shell, web, memory, and AWS capabilities out of the box.
openhands-sdk
Clean, modular Python SDK for building AI agents that work with code
OpenEvals
Prebuilt evaluators and LLM-as-judge scorers for testing the quality of LLM application outputs.
trustcall
Tenacious, validation-safe LLM tool calling and structured extraction built on LangGraph.
Stagehand (Python)
AI browser automation for Python that mixes natural-language actions with precise code
Bashkit
A blazing-fast, sandboxed virtual bash interpreter and file system for running untrusted AI-agent shell commands safely in-process.
agent-client-protocol
The official Rust SDK for the Agent Client Protocol (ACP), the JSON-RPC standard for wiring code editors up to AI coding agents.
toolbox-core
Google's framework-agnostic Python SDK for loading and invoking MCP Toolbox database tools as native async functions.
langchain-experimental
A staging ground for LangChain's newest, riskiest agent and reasoning tools, including Tree of Thought, autonomous agents, and LLM-driven SQL and graph chains.
@microsoft/agents-activity
TypeScript types and Zod validators implementing Microsoft's Activity Protocol for building conversational agents.
arcade-js
Official TypeScript/JavaScript client for the Arcade API, giving agents typed access to tool execution, authorization, and chat completions.
langchain-tavily
The official LangChain integration for Tavily, exposing search, extract, crawl, map, and research as ready-to-use agent tools.
n8n
Fair-code workflow automation platform with native AI and 400+ integrations
Graft
A CLI that builds a persistent, readable context graph of your codebase so coding agents stop re-exploring it from scratch every session.
