Works out of the box
Windows, macOS, and Linux packages include the Hermes Agent core. Download, install, add an API key, and start chatting in 3 minutes — no command line, no complex setup.
v0.6.3 · Beta preview By the Hermes Agent Chinese Community
The native Hermes Agent
desktop client
A native desktop client for Windows, macOS, and Linux with the Hermes-CN-Core runtime built in. Install it, set an API key or local model endpoint, and start working with your AI agent — chat, skills, memory, scheduled tasks, and MCP, all included.
Windows · macOS · Linux · Tauri v2 · Built-in Hermes-CN-Core
Hermes Agent Chinese Community · Official Built on a modern, native stack
- Tauri v2
- Rust
- React
- Windows
- macOS
- Linux
Why this app
What a desktop client should be
Lightweight & native
Built with Tauri. A small installer, fast startup, native windows, local processes, and system file dialogs.
Production-ready
A built-in independent Hermes Agent core with Ed25519 signature verification, health checks, and one-click rollback. REST, uploads, and SSE all go through a Rust proxy.
Live demo
See the desktop app in a minute
From installation and model setup to chat and workflows — the full agent experience, right on your desktop.
Features
A complete workbench for agents
Chat, streaming output, attachments, MCP, Skills, Memory, Profiles, scheduled tasks, and runtime health panels are all built in.
Chat & workflows
A complete chat experience built for agent workflows.
- Streaming chat Real-time streaming output with multi-turn context — responses appear instantly.
- Conversation history Full session history; revisit and resume unfinished tasks anytime.
- Attachments Drag files straight into the conversation for the agent to handle.
- LaTeX & Mermaid Native rendering for LaTeX, Mermaid, and similar rich content, so formulas, diagrams, and structured notes display directly.
Agent capabilities
The core capabilities that let an agent actually get work done.
- Skills, Memory & MCP Manage built-in skills, long-term memory, and MCP tool services in one place, with the agent context configured centrally.
- Profiles Multiple configuration profiles, switchable in one click.
- Scheduled tasks Describe it in natural language to create a scheduled job.
Models & deployment
Compatible with cloud providers and local model deployments.
- Cloud and local models Native support for Volcengine ARK, Qianfan, Hunyuan, Bailian, Zhipu GLM, MiniMax, Kimi, plus local Ollama, vLLM, LM Studio, and llama.cpp deployments.
- Model configuration Manage keys and parameters across providers, switch anytime.
- Credentials & OAuth OAuth sign-in and centralized credential management.
Runtime & security
Production-grade stability and security, handled by the desktop.
- Managed runtime A bundled runtime for first launch, with download, update, and one-click rollback for upgrades or repair.
- Signature verification Ed25519 verification ensures the core comes from a trusted source.
- Diagnostics & logs Dashboard, gateway, runtime health status, and logs — issues at a glance.
Model ecosystem
Connect the models you already use
Cloud providers and local deployment options such as Ollama, vLLM, LM Studio, and llama.cpp are covered — just add an API key or local endpoint and go.
- Volcengine ARK
- Baidu Qianfan
- Tencent Hunyuan
- Alibaba Cloud Bailian
- Zhipu GLM
- MiniMax
- Moonshot Kimi
- Xiaomi MiMo
- StepFun
- OpenRouter
- SiliconFlow
- ModelScope
- UCloud AI Computing
- Ollama local
- vLLM local
- LM Studio local
- llama.cpp local
A look inside
Crafted for productivity, down to the detail
Workbench, archives, chat, LaTeX, console, Feishu integration, stats, health, config, runtime, and logs.
Get started
Download Hermes Agent CN Desktop
Completely free and fully open source. Currently a Beta preview.
v0.6.3 Beta stage: features and packaging are still iterating — feedback welcome.
What's new in this release
- New Linux x64 desktop build with both a .deb installer and portable AppImage
- Linux bundles runtime-v0.18.2-cn.2 and can start the local core immediately
- Windows and macOS packages updated to v0.6.3
- Fixed completed turns being rendered as duplicate bubbles
Installer
Double-click to install, with update prompts — right for most users.
For your system
Windows
x64 · Windows 10 / 11
Download .exeHermes.Agent.CN.Desktop_0.6.3_x64-setup.exe
For your system
macOS
Apple Silicon（M 系列芯片用这个）· macOS 11+
Download .dmgHermes.Agent.CN.Desktop_0.6.3_aarch64.dmg
For your system
Linux
x64 · Ubuntu 22.04+ / Debian 12+
Download .debHermes.Agent.CN.Desktop_0.6.3_amd64.deb
Portable (no install)
Windows and macOS portable zips keep data in an adjacent data folder. The Linux AppImage runs without installation and uses the standard Linux user data directory.
macOS
Apple Silicon（M 系列芯片用这个）· macOS 11+
Download .zipHermes.Agent.CN.Desktop_0.6.3_aarch64-macos-portable.zip
Linux
x64 · Ubuntu 22.04+ / Debian 12+
Download .AppImageHermes.Agent.CN.Desktop_0.6.3_amd64.AppImage
Windows, macOS, and Linux packages include the Hermes-CN-Core runtime. On first launch, the app initializes from the bundled runtime first; managed download/update is used for upgrades or fallback repair.
First launch on macOS: right-click the app → “Open”, or allow it under System Settings → Privacy & Security.
See all versions & changelog → Community
Join the Hermes Agent Chinese Community
Explore real-world industry applications of AI Agent technology.
WeChat
68 active groups · daily digest & Q&A
Scan to join 
Feishu
5,000+ member group
Scan to join 
QQ
3,000+ member group
Scan to join The WeChat group QR rotates — click for the latest entry See all community entries
FAQ
Things you might wonder
How is this different from the web version / upstream?
This project is the desktop shell for Hermes Agent, adding native windows, local process management, system file dialogs, a managed runtime, runtime diagnostics, and a safer REST / SSE proxy layer in production mode. The agent runtime and Dashboard source live in Hermes-CN-Core; upstream is the Nous Research Hermes open-source project.
Is my data safe?
The desktop app runs the core and session state locally, and model calls go directly to the providers you configure. In production mode all REST, upload, and SSE traffic goes through a local Rust proxy that centralizes auth — never through a third-party server.
Which models are supported?
Native support for Volcengine ARK, Baidu Qianfan, Tencent Hunyuan, Alibaba Cloud Bailian, Zhipu GLM, MiniMax, Moonshot Kimi, SiliconFlow, ModelScope, UCloud AI Computing, plus local Ollama, vLLM, LM Studio, and llama.cpp deployments; other compatible services can be configured too.
Linux or Intel Macs?
Today we ship Windows x64, macOS Apple Silicon, and macOS Intel builds. Linux remains on the roadmap — watch GitHub Releases for future builds.
Open source & upstream
Built in the open, credit to upstream
The desktop app is developed in the open on GitHub. Its core comes from the Nous Research Hermes project, maintained and enhanced by the Chinese community. Star it, file issues, contribute.