跳到正文
Hatcher网站

框架

选择合适的框架

Hatcher 运行三个可用于生产环境的智能体框架 — OpenClaw 适合技能密集型自动化,Hermes 适合长期运行的自适应智能体,IronClaw 则为注重隐私的 Rust 原生智能体而生。

快速概览

OpenClaw

Advanced

Skill-heavy autonomous worker

Power users who need maximum capability: autonomous task execution, persistent workspace, skills, cron, and multi-channel integrations.

2,500+ ClawHub skills
Browser automation
Cron jobs & triggers
文档

Hermes

Intermediate

Learning agent

Agents that learn and adapt: long-running memory, live config editing, tool routing, and persistent research workflows.

2,500+ ClawHub skills
Browser automation
Cron jobs & triggers
文档

IronClaw

Intermediate

Rust-native private runtime

Privacy-focused agents on a hardened Rust runtime: WASM-sandboxed extensions, private libSQL memory, per-tool permissions, and conversational automations.

Cron jobs & triggers
Persistent memory
Plugin ecosystem
文档

功能矩阵

功能OC OpenClawHM HermesIC IronClaw
2,500+ ClawHub skills
Browser automation
Cron jobs & triggers
Multi-channel gateway
Persistent memory
Character personas
Plugin ecosystem
Social media native
Lightweight deploy

最适合...

不确定选哪个框架?根据使用场景进行匹配。

OC

OpenClaw

Advanced

Power users who need maximum capability: autonomous task execution, persistent workspace, skills, cron, and multi-channel integrations.

Enterprise automation
Multi-step research
Code generation pipelines
Cron-driven back-office agents
Full multi-platform bots
TelegramDiscordSlackWhatsAppX (Twitter)
HM

Hermes

Intermediate

Agents that learn and adapt: long-running memory, live config editing, tool routing, and persistent research workflows.

Research assistants
Knowledge base bots
DevOps copilots
Security/red-team helpers
Long-running memory workflows
TelegramDiscordSlackSignalMatrixEmail/SMS
IC

IronClaw

Intermediate

Privacy-focused agents on a hardened Rust runtime: WASM-sandboxed extensions, private libSQL memory, per-tool permissions, and conversational automations.

Privacy-sensitive assistants
Tool-permission-gated workflows
Scheduled native automations
Telegram-native agents
Encrypted credential handling
TelegramGitHubGmailNotionSlack

性能基准

在Hatcher共享基础设施上的冷启动时间和基准内存用量,实际值因配置而异。

冷启动时间

OC OpenClaw
3.2s
HM Hermes
2.8s
IC IronClaw
1.1s

越低越快

基准内存用量

OC OpenClaw
400 MB
HM Hermes
350 MB
IC IronClaw
120 MB

越低越轻量

OpenClaw 适合需要广泛技能和自动化能力的场景。Hermes 更轻量,适合带记忆和研究流程的自适应智能体。

配置示例

每个框架都有自己的配置格式,以下是典型设置示例。

# openclaw/config.yaml
model: deepseek/deepseek-v4-flash
provider: openrouter

skills:
  - web_search
  - calculator
  - file_manager
  - code_interpreter

platforms:
  telegram:
    token: "${TELEGRAM_BOT_TOKEN}"
  discord:
    token: "${DISCORD_BOT_TOKEN}"

memory:
  backend: sqlite
  path: ./workspace/memory.db
完整配置参考 →OpenClaw 文档

LLM 支持

两种框架都支持 BYOK,由用户直接向提供商付费。托管 LLM 优先使用 UsePod,并以 OpenRouter 作为 fallback,然后消耗 AI Credits。

OC OpenClaw

OpenRouter
OpenAI
Anthropic
Google
xAI
BYOK

HM Hermes

OpenRouter
OpenAI
Anthropic
Google
xAI
BYOK

IC IronClaw

OpenAI Responses API
NEAR AI

* BYOK 始终免费使用 Hatcher 平台;提供商用量由您自己的密钥支付。 查看 BYOK 设置指南 →

快速决策指南

我想要最多工具和自动化能力

OC · OpenClaw

我想要一个随时间学习的智能体

HM · Hermes

我需要多渠道消息和触发器

OC · OpenClaw

我需要一个重记忆的研究助手

HM · Hermes

准备好部署了吗?

选择框架,配置您的智能体,60秒内上线。免费套餐已包含。

Hatcher

选择用户名

为你的 Hatcher 账户设置专属名称。此名称会显示在个人资料和共享智能体中。

3–30 个字符:英文字母、数字、下划线或连字符。