Skip to Content
4회차1. 헤르메스 에이전트 소개

1. 헤르메스 에이전트 소개

한 줄 정의 (공식)

헤르메스 에이전트(Hermes Agent) — Nous Research가 공개한 오픈소스 에이전트. 공식 슬로건은 다음 한 문장입니다.

“The Agent That Grows With You.” 당신과 함께 자라는 에이전트.

hermes-agent.nousresearch.com

자기 정의 한 문단(공식 홈페이지, 직접 인용 + 한국어 번역):

“Not a coding copilot tethered to an IDE or a chatbot wrapper around a single API. An autonomous agent that lives on your server, remembers what it learns, and gets more capable the longer it runs.”

IDE에 묶인 코딩 copilot도 아니고, 단일 API를 감싼 챗봇 wrapper도 아닙니다. 본인 서버에 사는 자율 에이전트로서, 배운 것을 기억하고 오래 돌수록 더 유능해집니다.

공식 6가지 특징 (homepage 그대로)

영문 (원문)한국어 번역
Lives Where You Do본인이 머무는 곳에 산다 — Telegram · Discord · Slack · WhatsApp · Signal · Email · CLI 어디서든
Grows the Longer It Runs오래 돌수록 자란다“It learns your projects, auto-generates skills, and never forgets how it solved a problem.” / 본인 프로젝트를 학습하고, 스킬을 자동 생성하며, 한 번 풀어낸 문제는 잊지 않는다
Scheduled Automations자연어 스케줄 자동화“Natural-language scheduling for reports, backups, and briefings — running unattended through the gateway.” / 리포트·백업·브리핑을 자연어로 스케줄링하고, 게이트웨이를 통해 사람 손 없이 돌린다
Delegates & Parallelizes위임·병렬 실행“Isolated subagents with their own conversations, terminals, and Python RPC scripts.” / 각자 자기 대화·터미널·Python RPC 스크립트를 가진 격리된 서브에이전트
Real Sandboxing실 sandbox“local, Docker, SSH, Singularity, Modal — with container hardening and namespace isolation.” / 5종 백엔드(local · Docker · SSH · Singularity · Modal) + 컨테이너 하드닝과 네임스페이스 격리
Full Web & Browser Control전체 웹·브라우저 조작 — 브라우저 컨트롤과 vision까지

스킬 = 슬래시 커맨드 (핵심 개념)

공식 퀵스타트가 짚는 핵심 두 문장:

“Each is a SKILL.md file with a name, a description, and a step-by-step procedure.”

각 스킬은 이름·설명·단계별 절차가 적힌 SKILL.md 파일 한 장이다.

“Every installed skill becomes a slash command automatically: /k8s deploy

설치된 모든 스킬은 자동으로 슬래시 커맨드가 된다 — 예: /k8s deploy

스킬 한 장(SKILL.md markdown)이 곧 슬래시 커맨드로 등록되는 구조. 라이브 본 행사(3. 내 학습 키워드를 SKILL.md로)의 출발점이 바로 이 한 줄.

기본 정보

항목출처
메이커Nous Researchgithub
최초 공개2025-07-22github API
라이선스MITgithub
주 언어Pythongithub
모델 컨텍스트 요구64,000 토큰 이상quickstart
데스크탑 지원macOS 12+ · Windows 10·11 (Linux는 터미널 설치)/desktop

OpenClaw와 핵심 차이

헤르메스 = 자율 에이전트(autonomous agent), OpenClaw = 개인 어시스턴트(personal assistant).

가장 큰 차이는 주도권이 어디 있는가 — 헤르메스는 에이전트가 능동적으로 일을 진행하는 쪽, OpenClaw는 사람이 부르면 응답하는 쪽. 두 프로젝트의 공식 자기 정의 한 문장에 그대로 박혀 있는 차이입니다.

세 측면을 공식 인용으로 비교

측면헤르메스 (공식)OpenClaw (공식)
정체성”An autonomous agent” — 자율 에이전트”A personal AI assistant” — 개인 어시스턴트
위치”lives on your server” — 본인 서버에 산다”you run on your own devices” — 본인 디바이스에서 돌린다
행동 양식”gets more capable the longer it runs” — 오래 돌수록 자란다”answers you on the channels you already use” — 본인이 평소 쓰는 채널에서 응답한다

사용 시나리오로 풀어보면

  • 헤르메스적 사용“매일 아침 7시에 어제 outbox를 분석해서 한 줄로 알려달라” 라고 한 번 설정하면, 본인이 부르지 않아도 매일 아침 결과가 도착. “running unattended through the gateway.”
  • OpenClaw적 사용 — 본인이 Telegram·Slack 같은 채널에 메시지를 보내면 그때 응답. 호출이 없으면 잠잠. “It answers you on the channels you already use.”

같은 자료(SKILL.md·메모리·모델 provider)를 다루지만 주도권이 다릅니다 — 헤르메스는 에이전트가 주도, OpenClaw는 사람이 주도.

메타데이터·기능 세부 비교

항목헤르메스 에이전트OpenClaw
메이커Nous ResearchPeter Steinberger (steipete)
라이선스MITMIT
주 언어PythonTypeScript / Node.js (22.19+ 또는 24)
첫 공개2025-07-22OpenClaw 이전부터 운영
채널 통합Telegram · Discord · Slack · WhatsApp · Signal · Email · CLI20+ 채널 (WhatsApp · Telegram · Slack · Discord · Google Chat · Signal · iMessage · IRC · Microsoft Teams · Matrix · Feishu · LINE · Mattermost · Nextcloud Talk · Nostr · Synology Chat · Tlon · Twitch · Zalo · WeChat · QQ · WebChat 등)
핵심 구조 강조스킬SKILL.md 한 장이 곧 슬래시 커맨드게이트웨이“The Gateway is just the control plane — the product is the assistant.” / 게이트웨이는 컨트롤 플레인일 뿐, 제품은 어시스턴트 그 자체다.
데스크탑/모바일 부가 기능5 백엔드 sandbox · Python RPC subagentlive Canvas · macOS/iOS/Android 음성 입출력

4회차에서는 헤르메스를 다루지만 OpenClaw도 같은 영역(로컬-퍼스트 · MIT · 다채널 통합)에 있는 선택지라는 점만 짚어둡니다. 어느 쪽이 본인에게 맞는지는 라이브에서 본인이 직접 한 번 셋팅해보고 판단합니다.

출처 (전부 공식)

→ 다음: 2. 로컬 셋팅 한 단계씩

Last updated on