Skip to main content
Herm is Managed Hermes Agents: an API for deploying customer-facing AI agents built on the open-source Hermes harness. Bring your system prompt, skills, MCP tools, and connectors — each deployment gives you an isolated, container-sandboxed agent with memory, automations, self-learning, and a persistent filesystem, plus an endpoint to chat with it over SSE.

​
How it works

  1. Create a deployment for each of your customers with your system prompt, skills, MCP servers, and connector secrets.
  2. Send messages to the deployment’s default thread, or create new threads per conversation.
  3. Stream events over SSE to render tokens, tool calls, and status updates in your UI.

​
What you can do

Create Deployment

Deploy a dedicated Hermes agent for a customer in one API call.

Get Deployment

Fetch a deployment’s status, model, and filesystem details.

List Deployments

Browse all deployments across your customers.

Delete Deployment

Tear down an agent and its container.

Send Message

Send a message to a deployment’s agent on a thread.

Stream Events

Subscribe to tokens, tool calls, and status updates over SSE.

​
Base URL

All API requests use the following base URL: 
https://api.prismagents.com
Authenticate with a Bearer token — see Authentication for details. Herm is in early access; book a demo to get an API key.