C

CrewAI

agent_platform 官网 ↗ 官网 ↗ 数据已核验 2026-11-17

📊 评分

编程能力
8.5
性价比
9.0
灵活性
9.5
国内体验
7.5

🚀 快速开始

安装步骤

  1. 1 安装 Python >=3.10 且 <3.14 和 uv,运行 uv tool install crewai
  2. 2 运行 crewai create crew my_project;当前模板使用 agents/*.jsonc 与 crew.jsonc,不再按旧 YAML 模板描述
  3. 3 配置模型凭据和工具,执行 crewai install,再 crewai run
  4. 4 按需检查 telemetry / share_crew、人工审批和持久化配置
注册 / 获取账号 ↗
🇨🇳 国内使用提示: 可本地运行并选择提供商;安装/模型服务在国内的可达性未实测,需自行评估地区政策。

💲 定价

方案 价格 限制
Self-hosted framework $0 MIT Python 框架;模型/算力另计 去官网 →
Cloud Basic $0 50 workflow executions/月、2 个 Automations;不是无限量云服务 去官网 →
Enterprise 定制定价;部署/治理/支持按合同 去官网 →

开源 Python 框架与托管 Control Plane / Enterprise 服务分开计费。默认匿名使用分析可用 OTEL_SDK_DISABLED=true 关闭;share_crew=True 会分享更详细的任务/输出数据。不能把 MIT 或自托管理解为没有外发、遥测和模型费用。 历史编辑评分沿用,未做新的实测或重评,不代表本次性能或性价比基准。

CrewAI

列有免费软件或档位;适用范围和注册条件见定价说明

官网 ↗

✨ 核心功能

代码补全 —
对话 ✅
Agent 模式 ✅
多文件编辑 —
MCP 支持 ✅
Sub-agent ✅
多 Agent ✅
agent_teams ✅
task_management ✅
persistent_memory ✅
skill_sharing ✅
runtime_monitoring ✅
open_source ✅
self_hosted ✅
role_based ✅
event_driven ✅
human_in_the_loop ✅
上下文窗口 取决于所选模型
supported_runtimes 内置 Agent 引擎(Python)
supported_os macOS, Windows, Linux
ide_base Agent 协作平台

🎯 最适合

Python 开发者构建多 Agent 自动化需要精确工作流控制的企业场景角色扮演式 Agent 协作(Crews)

📝 更新记录

2026-10-13 Version bump 1.14.4 → 1.15.27 (verified via GitHub releases)
2026-05-09 加入 CodePick Agent 协作栏目

📚 相关指南

🔗 相关工具

💬 用户评论

最后验证: 2026-11-17 下次评审: 2026-12-17 数据信心: medium 报告错误 →