CubeLoop — a Pythonic, async-native agent harness core designed for high performance, readability, and production-grade persistence.
Python agent-frameworkagent-harnessagent-runtimeagentic-aiai-agentsai-frameworkanthropicharnessharness-engineeringharness-frameworklanggraphlanggraph-alternativellmmcpopenaipi-agentpi-agent-corepythonpython-agent
import React from 'react';
import Link from '@docusaurus/Link';
import { useIsZhHans } from '@site/src/hooks/useIsZhHans';
import styles from './WhyTable.module.css';
const ROWS_EN: { label: string; langgraph: string; cubeloop: string }[] = [
{ label: 'Abstraction', langgraph: 'Graph nodes + edges + channels', cubeloop: 'Plain async functions — run_agent_loop is a while loop' },
{ label: 'Streaming', langgraph: 'Callback-based, multiple handler types', cubeloop: 'async for event in stream — one pattern everywhere' },
{ label: 'Checkpointing', langgraph: 'Full snapshot per step; serializes entire message list', cubeloop: 'Append-only — O(1) DB I/O regardless of conversation length' },
{ label: 'Dependencies', langgraph: 'langchain-core, langgraph-sdk, and transitive deps', cubeloop: '3 core deps: pydantic, anthropic, openai' },
{ label: 'Tool execution', langgraph: 'Tools are graph nodes with manual wiring', cubeloop: 'Declare tools as functions; framework routes and parallelizes' },
{ label: 'Multi-provider', langgraph: 'Via langchain chat model adapters', cubeloop: 'Native Provider protocol — Anthropic, OpenAI built in' },
{ label: 'Middleware', langgraph: 'Graph-level middleware on node entry/exit', cubeloop: '8 typed hooks with declarative composition rules' },
{ label: 'Observability', langgraph: 'LangSmith / Langfuse integration', cubeloop: 'Native OpenTelemetry — Tracer, Meter, GenAI semconv, OTLP / JSONL out of the box' },
];
const ROWS_ZH: { label: string; langgraph: string; cubeloop: string }[] = [
{ label: '抽象模型', langgraph: '图节点 + 边 + 通道', cubeloop: '普通 async 函数 — run_agent_loop 就是一个 while 循环' },
{ label: '流式输出', langgraph: '基于回调,多种 handler 类型', cubeloop: 'async for event in stream — 统一模式' },
{ label: '检查点', langgraph: '每步全量快照,序列化整个消息列表', cubeloop: '追加式 — 无论对话多长,每轮 O(1) DB I/O' },
{ label: '依赖项', langgraph: 'langchain-core、langgraph-sdk 及传递依赖', cubeloop: '3 个核心依赖:pydantic、anthropic、openai' },
{ label: '工具执行', langgraph: '工具是需要手动连线的图节点', cubeloop: '声明为函数;框架自动路由并并行执行' },
{ label: '多 Provider', langgraph: '通过 langchain chat model 适配器', cubeloop: '原生 Provider 协议 — 内置 Anthropic、OpenAI' },
{ label: 'Middleware', langgraph: '图级中间件,在节点进出时触发', cubeloop: '8 种类型化 hook,声明式组合规则' },
{ label: '可观测性', langgraph: 'LangSmith / Langfuse 集成', cubeloop: '原生 OpenTelemetry — Tracer、Meter、GenAI semconv,开箱即用 OTLP / JSONL' },
];
export default function WhyTable() {
const zh = useIsZhHans();
const ROWS = zh ? ROWS_ZH : ROWS_EN;
return (
<section className={styles.section}>
<h2 className={styles.h2}>
{zh ? '为什么选择 CubeLoop — langgraph 和 pi-agent-core 的替代方案' : 'Why CubeLoop — a langgraph and pi-agent-core alternative'}
</h2>
<div className={styles.tableWrap}>
<table className={styles.table}>
<thead>
<tr>
<th></th>
<th>langgraph</th>
<th>CubeLoop</th>
</tr>
</thead>
<tbody>
{ROWS.map((r) => (
<tr key={r.label}>
<td className={styles.label}>{r.label}</td>
<td>{r.langgraph}</td>
<td className={styles.us}>{r.cubeloop}</td>
</tr>
))}
</tbody>
</table>
</div>
<p className={styles.more}>
<Link to="/compare/langgraph">
{zh ? '完整对比:CubeLoop vs LangGraph →' : 'Full comparison: CubeLoop vs LangGraph →'}
</Link>
{' · '}
<Link to="/compare/pi-agent-core">
{zh ? 'CubeLoop vs pi-agent-core →' : 'CubeLoop vs pi-agent-core →'}
</Link>
</p>
</section>
);
}