The Platform
Hermes, Without the Ops
Hermes is a full autonomous agent framework by Nous Research (128K+ GitHub stars, MIT license). It doesn't just respond to messages — it maintains persistent memory of your work, schedules its own tasks, executes composable skill packs, and runs 24/7 without needing to be asked. Vessel manages the infrastructure so you never have to.
What Hermes is
Hermes is not a chat assistant you talk to. It's an autonomous agent framework — a persistent process that maintains a model of your work over time, executes structured workflows on its own schedule, and surfaces results to you through the channels you already use.
The distinction matters. A reactive agent waits for your message and responds. An autonomous agent has a task queue, a memory of what you care about, and a schedule. It completes work whether you asked for it today or set it up last week.
Persistent memory
Hermes maintains a structured memory of your preferences, context, and corrections across every session. It builds a user profile that sharpens its judgment over time — injected into the system prompt automatically on every turn.
Native cron scheduling
Hermes has a built-in cron subsystem. You schedule recurring tasks by chatting with the agent (“check our metrics every Monday at 9am”) and it manages its own schedule. Jobs survive container restarts and run without you being online.
Composable skill packs
25+ built-in skill packs across devops, research, email, GitHub, productivity, social media, note-taking, and more. Skills are modular — enable the ones your agent needs, leave the rest off. No rebuild required.
14+ messaging channels
Telegram, Discord, Slack active now. WhatsApp, Signal, Matrix, Email, Mattermost available. All channels share the same persistent memory and task queue — one agent, every surface.
What your agent can do from the first message
Every Hermes vessel ships with a full autonomous tool set enabled by default on all messaging channels. No plugins to install, no permissions to configure.
Built-in tools (active on every channel, no setup)
Terminal access
Execute shell commands, install packages, run scripts. The agent has a full Linux environment and uses it to complete real tasks.
File read & write
Create, edit, and read documents in the agent's workspace. Markdown, code, plain text — whatever the task requires.
Browser automation
Open pages, interact with sites, take screenshots via Playwright. Full browser control, not just URL fetching.
Web content fetch
Download and parse web pages for research, monitoring, and data extraction. Works alongside full browser access.
Cron job management
Schedule recurring tasks by chatting. "Send me a summary every Friday at 5pm" becomes a persistent job — no code, no config.
Persistent memory
Structured notes about your preferences, context, and corrections injected on every session. The agent learns your work.
Image generation
Generate images from text prompts directly in chat. Useful for content, design references, and visual output.
Vision
Analyze images and screenshots. The agent can describe, extract, and reason about visual content sent through any channel.
Task list (todo)
Structured task tracking built into the agent. Persists across sessions and is surfaced through your messaging channel.
Optional (bring your own key)
Text-to-speech
Have the agent deliver audio output via ElevenLabs. Useful for hands-free summaries and accessibility workflows.
OpenAI-compatible REST API
Every Hermes vessel exposes an OpenAI-compatible API endpoint. Call your agent programmatically from other systems using standard tools — no custom SDK needed.
Hermes or OpenClaw?
Both runtimes run on dedicated private VMs. The choice is about how your agent works, not where it runs.
| Capability | OpenClaw | Hermes |
|---|---|---|
| Messaging channels | 20+ | 14+ |
| Persistent memory | File-based | Structured (profile + notes) |
| Cron / scheduled tasks | Via skills | Native, chat-managed |
| Skill packs | Thousands (ClawHub) | 25+ built-in packs |
| REST API endpoint | — | OpenAI-compatible |
| Reactive (responds to messages) | ✓ | ✓ |
| Autonomous (works without being asked) | Partial | ✓ |
| Best for | Messaging-first agents | Structured autonomous work |
If you want an agent that handles your WhatsApp, Slack, and email simultaneously — drawing on a large library of community skills — that's OpenClaw. If you want an agent that maintains a structured model of your work, schedules its own recurring tasks, and operates with minimal prompting — that's Hermes.
The difference is not features. It's posture. OpenClaw waits for you. Hermes works while you're not looking.
What “without the Ops” actually means
Hermes is powerful precisely because it's a full framework — and full frameworks have real infrastructure requirements. Docker, compose files, data volumes, networking, secrets management, health monitoring, and channel credential rotation. Most people who would benefit most from an autonomous agent shouldn't need to know what any of that means.
Every vessel is a dedicated virtual machine on Google Cloud with Hermes pre-configured and running. You provision from the dashboard. Your agent is running in minutes.
| Task | Self-hosted | Vessel |
|---|---|---|
| Provision a server | You | Managed |
| Install & configure Hermes | You | Managed |
| Set up networking & tunnels | You | Managed |
| Manage data volumes & secrets | You | Managed |
| Monitor health & auto-restart | You | Managed |
| Apply updates | You | Managed |
| Security hardening | You | Managed |
| Choose your LLM provider | You | You |
| Train the agent on your work | You | You |
| Configure your skill packs | You | You |
Inside your vessel
When you provision a Hermes vessel: a dedicated GCP virtual machine spins up, Hermes starts in a Docker container, a private encrypted tunnel connects it to your messaging channels, and your access token is generated. The data volume persists your memory and scheduled tasks across every restart.
Your vessel / architecture
Built for work that runs without you
Hermes is not for developers who want to experiment with agent infrastructure. It's for professionals who need an agent that executes structured, recurring work — without being managed like a junior employee who needs constant direction.
If you need someone to draft your weekly client summary every Friday, monitor your competitors every morning, or process a document queue while you sleep — Hermes does that work. You check the output. You correct the judgment calls. The agent runs the rest.
Recurring research
Schedule morning competitive briefings, weekly market reports, or regulatory monitoring. The agent runs on schedule and delivers to your Slack or Telegram.
Document processing queues
Drop contracts, reports, or briefs into the agent's workspace. It processes them, extracts what matters, and surfaces the output — on your timeline, not its.
Long-horizon projects
Tasks that take days or weeks: due diligence packages, audit trails, research compilations. The agent tracks progress, maintains context, and picks up where it left off.
Programmatic access
Every Hermes vessel exposes an OpenAI-compatible REST API. Call your agent from other systems, pipelines, or tools using standard HTTP — no custom integration required.
128K+
GitHub stars
Nous Research open-source
MIT
License
Free, open, no lock-in
14+
Messaging channels
Telegram, Discord, Slack, and more
Not sure which runtime fits your work? Read the OpenClaw page and compare — the choice becomes obvious once you know what kind of work you need the agent to do.
Both runtimes run on the same isolated private VM infrastructure. Same security posture. Same provisioning flow. See how provisioning works.
