Open-source AI agents
Hermes Agent vs OpenClaude
OpenClaude is an open-source coding agent for the terminal, built from Claude Code and opened up to many models. Hermes Agent is an always-on assistant that lives in your chat apps. Some people who search “OpenClaude” actually mean OpenClaw. Here is how to tell which one you need.
Short answer: Choose OpenClaude to write and fix code in a terminal with the model you pick. Choose Hermes Agent for an assistant that keeps running, remembers context and reports back in Telegram, Slack or WhatsApp. If you were looking for OpenClaw, the agent framework, see Hermes Agent vs OpenClaw.
How we wrote this: We run OpenClaw Launch, which hosts Hermes Agent and OpenClaw and ships OpenClaude pre-installed in our Coding Harness, so we have a stake in all three. OpenClaude details were checked against its GitHub README and npm package on October 3, 2026. Searches for “openclaude” rose about 180% on Google Trends in the 24 hours to October 1, 2026, and Google’s related queries were “opencode” and “openclaw”. Some searchers want a coding agent, and some are mixing up names.
What is OpenClaude?
OpenClaude is an independent, open-source coding-agent CLI from Gitlawb. It originated from the Claude Code codebase and has been substantially modified so one terminal workflow works across providers: OpenAI-compatible APIs, Gemini, GitHub Models, Codex, local Ollama models and other supported backends. Source: OpenClaude on GitHub.
It keeps the familiar coding-agent tools: shell commands, file editing, grep and glob, MCP, web tools and slash commands. You can resume or fork a conversation, run long prompts as background sessions, and launch it from a bundled VS Code extension. Its README states that it is not affiliated with, endorsed by or sponsored by Anthropic.
To install it yourself, use the scoped npm package @gitlawb/openclaude with Node.js 22 or later. The unscoped openclaude package on npm is an unrelated placeholder, so check the name before you install anything you will give API keys to.
What is Hermes Agent?
Hermes Agent is an open-source agent from Nous Research. It keeps memory across sessions, writes reusable skills from work it has done, runs scheduled jobs and talks to you through messaging apps. It works with the model of your choice, Claude included. Source: Hermes Agent on GitHub.
You can run Hermes on your own server or deploy it on OpenClaw Launch, where we manage the container and you connect Telegram, Discord, Slack, WhatsApp, WeChat or Feishu from the dashboard.
Quick comparison
| Category | OpenClaude | Hermes Agent |
|---|---|---|
| What it is | An open-source coding-agent CLI from Gitlawb, derived from the Claude Code codebase | An open-source agent from Nous Research that builds memory and skills as you use it |
| Built for | Reading and changing code in a repository from your terminal | Ongoing assistant work: research, summaries, scheduled jobs and chat |
| Where you use it | A terminal, plus a bundled VS Code extension | Telegram, Discord, Slack, WhatsApp, WeChat, Feishu and web chat on OpenClaw Launch; more adapters upstream |
| Models | OpenAI-compatible APIs, Gemini, GitHub Models, Codex, Ollama and other supported backends | Any supported provider, including Anthropic Claude with an API key |
| When it runs | While a session or background job you started is running | Always on, with a built-in scheduler that reports back in chat |
| Memory | Conversation sessions you can resume or fork | Persistent memory, session search and skills it writes itself |
| Tools | Shell, file editing, grep and glob, MCP, web tools and slash commands | Browser, terminal, files, MCP servers and skills |
| Relationship to Anthropic | Independent community project; not affiliated with or endorsed by Anthropic | Independent project from Nous Research; can call Claude models |
| On OpenClaw Launch | Pre-installed in the Coding Harness with a browser terminal | Managed Hermes Agent hosting |
Sources: OpenClaude, Hermes Agent, and our Coding Harness and integrations pages.
Coding sessions or an assistant that stays on
OpenClaude works the way Claude Code does: you open a terminal in a project, describe a change, and it reads files, edits them and runs commands until the job is done. It works while the session or background job you started is running. That is the right shape for fixing a bug, writing tests or refactoring a module.
Hermes is built for work that continues after you close the laptop. You message it in a chat app, it remembers your projects, and its scheduler can send you a morning summary or check something every hour. Hermes can run code too, but its strength is continuity rather than a long coding session in one repository. See Hermes Agent memory.
Models: both let you choose
Despite the name, OpenClaude is not limited to Claude. Its point is one coding workflow across many providers, including local models through Ollama. Hermes also lets you pick the model and switch it later, from a cheap model for daily digests to a stronger one for harder work. Either way, check that the model you choose handles tool calls reliably before you trust it with real changes.
If you want Claude in Hermes, use an Anthropic API key or the AI credits included in our plans. Anthropic does not permit third-party services to route requests through Claude Pro or Max subscription credentials. See why we don’t offer Claude subscription login and the model catalog.
If you meant OpenClaw
OpenClaw is a different project: an open-source agent framework from the OpenClaw Foundation that connects a model to 20+ messaging channels, tools and companion devices, with workspace memory and skills. Source: OpenClaw on GitHub. It competes with Hermes Agent, not with OpenClaude. Our Hermes Agent vs OpenClaw comparison covers that choice in detail.
Running them on OpenClaw Launch
Hermes Agent, OpenClaw and the Coding Harness use the same plans. The Coding Harness comes with OpenClaude and seven other coding agents pre-installed, a browser terminal and a persisted /workspace folder, so you can try OpenClaude without installing anything locally. OpenClaude signs in with your own provider key or login, exactly as it would on your laptop; the workspace does not hand it our AI credits or your saved keys.
- Free trial: try it before you subscribe, no credit card required.
- Lite: $3 for the first month, then $6/month, with $1/month in AI credits.
- Pro: $20/month, with $10/month in AI credits and up to 3 instances.
Hosting and AI usage are separate budgets. The included AI credits are for Hermes Agent and OpenClaw; heavy use beyond them needs a top-up or your own provider key. OpenClaude and Hermes are both free to run on your own machine; then you pay only your model provider and your own hardware.
Which should you choose?
Choose OpenClaude if you want a Claude Code-style coding agent in your terminal and want to choose the model, including local ones.
Choose Hermes Agent if you want an assistant that remembers your context, writes its own skills and runs scheduled jobs that report back in Telegram, Slack, WhatsApp or WeChat.
Choose OpenClaw if “OpenClaude” was a typo and you want the agent framework with the widest set of messaging channels and companion devices.
They are not exclusive. A common setup is Hermes for daily assistant work and a coding agent such as OpenClaude for sessions in a repository.
Frequently asked questions
What is OpenClaude?
OpenClaude is an open-source coding-agent CLI maintained by Gitlawb. It started from the Claude Code codebase and was changed to work with many model providers, including OpenAI-compatible APIs, Gemini, GitHub Models, Codex and local Ollama models. You run it in a terminal to read, edit and test code.
Is OpenClaude made by Anthropic?
No. The project describes itself as an independent community project that is not affiliated with, endorsed by or sponsored by Anthropic. “Claude” and “Claude Code” are Anthropic trademarks. Anthropic’s own coding agent is Claude Code.
Is OpenClaude the same as OpenClaw?
No. OpenClaude is a coding CLI. OpenClaw is an open-source agent framework from the OpenClaw Foundation that connects a model to messaging apps, tools and devices. The names are close, so some people type one when they mean the other.
How do I install OpenClaude?
The project publishes it on npm as @gitlawb/openclaude and needs Node.js 22 or later. Install from that scoped name. The bare npm name “openclaude” is a different, placeholder package, not the CLI.
Can I run OpenClaude on OpenClaw Launch?
Yes. Our Coding Harness is a managed cloud dev box with OpenClaude pre-installed alongside Claude Code, Codex, OpenCode, Aider, Zero, Pi and DeepSeek Harness. You open a browser terminal, sign in with your own model provider key or login as you would locally, and run it against a persisted /workspace folder.
Should I pick Hermes Agent or OpenClaude?
They do different jobs. Pick OpenClaude to work on code in a terminal with the model you choose. Pick Hermes Agent for an assistant that stays on, remembers context and reports to you in Telegram, Slack or WhatsApp. Many people use both.
Related comparisons and guides
- Coding Harness hosting: OpenClaude and seven other coding agents
- Hermes Agent vs Claude Code: the project OpenClaude came from
- Hermes Agent vs OpenClaw: if you meant OpenClaw
- OpenClaw Launch vs OpenCode: another open-source coding agent