Agent platform and harness

The open source AI agent platform and harness

A harness with persistent memory, a skills system and real tool use. Run the same agent infrastructure on your own machine or as a managed cloud runtime.

Start Plus, $20/mo Compare plans

Install via terminal

curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash

Message it from

Telegram
Discord
Slack
WhatsApp
Email
CLI

and more

Persistent memory

Saves useful details and writes reusable skills you can review.

Agent infrastructure you control

The same agent runs on your machine or in Nous Portal.

Tools and MCP

60+ built-in tools, plus any MCP server for more.

Delegate to subagents

Child agents work concurrently; your chat gets their summaries.

Inside the AI agent harness

The harness is the part around the model: what it remembers, what it can reach, and where its work actually runs.

#1 Runtime

The agent runtime

Run it in your terminal on macOS, Windows or Linux, or deploy the same agent as an always-on cloud runtime.

#2 Memory

Memory and skills

Persistent memory across sessions, plus portable skills it loads on demand and writes as it learns. Review what it keeps.

#3 Tools

Real tool use

60+ built-in tools it can call, and any MCP server can add more without touching core code.

#4 Isolation

Optional isolated backends

Shell commands run on your machine by default. Docker, SSH, Modal, Daytona and Vercel Sandbox add optional isolation per session or agent.

#5 Background

Background agents

Delegate a big job with a handle and keep chatting, or schedule one in natural language or cron; it runs unattended.

#6 Delegate

Parallel subagents

Spawn child agents with their own conversation, terminal session and toolset. They work concurrently; only their final summary enters your context.

From install to the cloud

The agent harness is free and open source under the MIT license. Install it, connect a provider, then run it anywhere.

Install the harness
One line in the terminal on Mac, Windows or Linux, or Hermes Desktop for Mac and Windows.
Connect a model provider
Use your own key — Nous Portal, OpenRouter, OpenAI, Anthropic, Google or any OpenAI-compatible endpoint.
Run locally or deploy
The same AI agent runtime works on your machine or as an always-on cloud agent.

Hermes Desktop app: agent list and a chat where the agent writes release notes from recent commits — The same harness with a graphical interface: Hermes Desktop.

Source on GitHub Read the docs

Hundreds of models, one account

The live Nous Portal catalog: search by provider or model and compare prices per million tokens.

200+ models

Browse the full catalog

Plans and pricing

Hermes Agent is free and open source. Nous Portal adds models, hosted tools and Cloud hosting under one account: start with pay as you go, or pick a monthly plan with included usage.

One account, one bill
Use models from many providers, plus hosted tools, without a separate signup and API key for each.
Switch models per task
Pick another model with /model and keep working in the same chat.
Hosted tools on paid plans
Web search, image generation, text-to-speech and cloud browser tools run through Nous, counted in your plan's usage.

FreePlusSuperUltra

Feature	Free	PlusPopular	Super	Ultra
Price	$0, pay as you go	$20 / month	$100 / month	$200 / month
Best for	Trying Hermes on free models	Everyday personal use	Frequent research and coding	Heavy individual use
Usage	Free	Shared	5× Plus	10× Plus
Pay-as-you-go top-ups
Models	Free models only	Free and paid models	Free and paid models	Free and paid models
Hosted toolsWeb search, images, speech, cloud browser	—
Cloud hostingOne agent, up to this size	—	4 GB RAM, 25 GB disk	4 GB RAM, 100 GB disk	8 GB RAM, 100 GB disk
Rate limits	Standard	High	High	High
Start	Start free	Start Plus	Start Super	Start Ultra

Cancel anytime · Paid plans are billed monthly · Included usage has daily, weekly and monthly limits. Past them, switch to a free model, wait for the reset, or keep going on pay-as-you-go credits.