Hermes Agent
Hermes Agent is an autonomous agent system built by Nous Research, a renowned open-source AI collective. Featuring cross-session persistent memory and a self-improving learning loop, it can autonomously write and build up dedicated skills based on task requirements. Connecting to the OpenRely API links its cognitive core directly to world-leading foundation models.
✨ AI One-Click Setup
Copy the prompt below and send it to your terminal AI assistant to automatically complete environment checks and configuration:
🔑 Prerequisites
sk-...🚀 Steps
Install and Initialize Hermes Agent
Hermes Agent is built on a Python environment. It is recommended to use pip or the modern package manager uv for quick installation:
# 推荐通过 pip 安装官方 CLI
pip install hermes-agent
# 或者克隆官方源码安装
git clone https://github.com/NousResearch/hermes-agent.git
cd hermes-agent
pip install -e .Configure OpenRely Service Endpoint and API Key
Hermes Agent natively supports standard OpenAI-compatible protocols. You can quickly connect to OpenRely via environment variables or a configuration file:
Export the configuration parameters in your terminal session:
export OPENAI_BASE_URL="https://api.openrely.ai/v1"
export OPENAI_API_KEY="sk-your-openrely-api-key"
export HERMES_MODEL="claude"[!TIP] For long-term effect, you can append the commands above to
~/.zshrcor~/.bashrc.
Create or edit ~/.hermes/config.yaml in your user home directory:
provider: openai
api_base: "https://api.openrely.ai/v1"
api_key: "sk-your-openrely-api-key"
model: "gpt" # 或 claude、deepseek 等，具体标识前往定价中心查询
memory:
enabled: true
storage: localStart an Interactive Terminal Session
Run hermes chat to start the interactive agent with persistent memory:
hermes chatUpon entering for the first time, send it a complex comprehensive task, for example:
> 帮我分析当前项目中的性能瓶颈，并编写一个自动化测试脚本沉淀为你的长期技能Hermes Agent will perform contextual analysis using its long-term memory store and attempt to write a dedicated Skill to save into the local skill pool.
💡 Core Features
| Feature Module | Description |
|---|---|
| Self-Improving Skills | When executing complex workflows, the agent can autonomously write reusable Python skill scripts and archive them persistently |
| Long-Term Memory | Automatically extracts key facts, code architectures, and personal preferences from conversations, preventing context loss across sessions |
| Multi-Channel Gateway Support | Built-in gateway service that makes it easy to attach the same Agent to Telegram, Discord, or private enterprise groups |
| Model-Agnostic Design | Deeply compatible with the standard OpenAI protocol; seamlessly switch to any top-tier foundation model globally under the OpenRely ecosystem |
❓ FAQ
Encountering other invocation errors or billing questions? Please refer to the complete FAQ and Troubleshooting Guide.
Was this page helpful?
OpenClaw
OpenClaw is an open-source, self-hosted personal AI agent and gateway platform. It seamlessly connects multi-channel messages from Telegram, Discord, WhatsApp, and more to AI agents running on local hardware, supporting multi-agent cooperative routing, persistent memory, and autonomous tool calling, allowing developers to take full control of their private AI assistants.
Cherry Studio
Cherry Studio is a cross-platform desktop AI assistant client built for professional users. It features 30+ built-in industry agent templates, supports multi-model concurrent comparison, local knowledge bases, document parsing, and long-context reasoning. Connect to the OpenRely API to schedule top global LLMs in one stop.