Hermes
Is Hermes good for me?
A self-improving AI agent with memory, skills, scheduling, subagents, and multi-platform access.
Tool short answer
Hermes Agent is Nous Research's self-improving AI agent. It combines a terminal interface, persistent memory, session search, agent-curated skills, model/provider switching, scheduled automations, subagents, messaging gateways, and API access for technical users who want an inspectable agent workflow.
5-day hosted trial
$5 credit if you upgrade to self-hosting
Hermes trials currently use OpenRouter BYOK. Your key is injected during VPS setup, then removed from Convex after the server is created.
- Hermes Agent installed on an isolated trial VPS
- Trial provider configuration, currently OpenRouter BYOK
- Generated dashboard/API credentials where enabled
- HTTPS status/API/dashboard routes for supported surfaces
- Provisioning log and readiness tracking
- Access for 5 days
5 days, then the instance is deleted
What is Hermes?
Hermes is not just a hosted OpenAI-compatible endpoint. Upstream Hermes is an agent runtime and operating loop: chat in the terminal or through messaging platforms, choose models from multiple providers, let the agent persist useful memory, create and refine skills from experience, search past sessions, schedule recurring jobs, and delegate parallel work to subagents. Evaluate it when you want to know whether that learning loop and tool-rich agent workflow fit how you work.
Who it's for
- Developers evaluating self-improving agent workflows with memory, skills, and tools
- Founders comparing OpenClaw-style channel assistants with a Hermes learning loop
- Teams that need BYOK model routing, scheduled automations, subagents, and inspectable infrastructure
- Builders who want CLI, messaging, dashboard/API, and gateway surfaces on their own VPS
Who it's not for
- Non-technical users expecting a polished managed SaaS
- Teams that cannot bring an OpenRouter API key for the trial
- Production workloads that need SSO, backups, and monitoring from day one
Problems Hermes solves
- Agent workflows that forget useful context instead of building memory and skills over time
- No quick way to evaluate terminal, messaging, scheduler, subagent, and API surfaces together
- Unclear model/provider switching and gateway behavior before a real evaluation
- Provider secrets and agent configuration scattered across local experiments
Why people self-host Hermes
- Provider keys stay on your VPS after provisioning
- You control gateway access, logs, and upgrade timing
- The endpoint can be tested with standard OpenAI-compatible clients
- No need to accept a hosted agent platform before validating the runtime
Setup difficulty if you do it yourself
The hosted trial handles the Linux install, firewall, Caddy, systemd services, provider config, generated credentials, and supported dashboard/API routes. Production operation still needs backups, monitoring, and version management.
With the hosted trial: isitforme.tech handles the server, DNS, TLS, and tool-specific access flow. You get the access details for this tool when provisioning completes.
What isitforme.tech provisions for the trial
- Hermes Agent installed on an isolated trial VPS
- Trial provider configuration, currently OpenRouter BYOK
- Generated dashboard/API credentials where enabled
- HTTPS status/API/dashboard routes for supported surfaces
- Provisioning log and readiness tracking
- Access for 5 days
BYOK required
Hermes trials currently use OpenRouter BYOK. Your key is injected during VPS setup, then removed from Convex after the server is created.
Tool capabilities
- Terminal UI with slash commands and streaming tool output
- Persistent memory, session search, and agent-curated skills
- Messaging gateway for Telegram, Discord, Slack, WhatsApp, Signal, and related surfaces where configured
- Cron scheduling and unattended automations
- Subagents and Python/RPC tool workflows
- Model/provider switching plus OpenAI-compatible API/gateway surfaces
After the trial: your upgrade path
Keep the VPS, reproduce the cloud-init steps on your own server, or migrate the Hermes config and provider setup into your production infrastructure.
Questions about Hermes
The first hosted trial path supports OpenRouter only. That keeps setup predictable while still giving access to a broad model catalog.
Hermes exposes an OpenAI-compatible API server for chat completions. Use the API base URL and key shown in your dashboard.
Convex clears the BYOK config after the Hetzner server is created. The key remains only in the Hermes environment file on the trial VPS until the VPS is deleted.
The hosted trial may expose a status/API surface and, where enabled, a web dashboard. Hermes itself also supports CLI and messaging workflows; the trial page should label exactly which surfaces this provisioner enabled.
Ready to try Hermes?
Get a live Hermes instance in minutes with the access needed to evaluate it. $5 for 5 days.