Skip to main content
Hermes Agent is an AI runtime that gives your agent self-improvement capabilities, automatic skill creation, built-in memory, browser access, and code execution — all behind a multi-channel gateway.

​
What you get

CapabilityDescription
GatewayHTTP gateway that accepts webhooks from connected channels and routes them to the agent
Skill creationAgent can create, edit, and invoke reusable skills to extend its own capabilities
MemoryBuilt-in conversation memory with configurable retention
BrowserHeadless browser for web navigation and scraping
Code executionSandboxed code execution for computation tasks
Multi-channelSingle agent instance serves multiple channels simultaneously

​
Create a Hermes agent

oc agent create my-hermes --core hermes
This boots a sandbox with Hermes pre-installed, an LLM key configured, and the gateway running. An instance is created automatically.

​
Configuration

Hermes uses YAML configuration at ~/.hermes/config.yaml inside the sandbox. When you connect channels or install packages via oc agent, the platform updates this config and restarts the gateway. To inspect or edit manually: 
oc shell my-hermes
cat ~/.hermes/config.yaml

​
Gateway

The Hermes gateway listens for incoming webhooks and dispatches them to the agent. When you connect a channel, the platform:
  1. Registers the channel’s webhook with the provider
  2. Adds the channel config to ~/.hermes/config.yaml
  3. Restarts the gateway with hermes gateway run --replace
Gateway logs are at ~/.hermes/logs/gateway.log.

​
Skills

Hermes can create skills — reusable functions it builds and invokes on its own. To see what skills an agent has created: 
oc shell my-hermes
hermes skill list

​
Extending with packages

Hermes supports packages that add capabilities via MCP servers. When a package is installed, the platform adds it to the Hermes MCP config and restarts the gateway.

Create a Hermes Agent

Step-by-step guide

oc agent CLI

CLI commands for managing agents