Menu

Hermes Agent icon

Hermes Agent

Hermes Agent

Hermes Agent is a Nous Research agent for tasks that need tools and continuity across sessions. Its documented capabilities include persistent memory, reusable skills, terminal and file tools, messaging gateways, and scheduled tasks. Evaluate it for repeated research or file workflows that need more than a single chat response. Official project.

Establish one working provider first

Follow the official quickstart, run hermes setup, and select a model provider with hermes model. Confirm that a normal conversation succeeds before adding a gateway or schedule.

Connect only the tools the task needs. Then try a bounded workflow, inspect the files or results, and check how a later session resumes it. Messaging integrations require separate platform configuration and access controls. Scheduling needs a running environment that can execute the task at the intended time.

Software and running costs

The repository uses the MIT license. You can run the software yourself, but model inference, search or browser services, and the host machine have their own costs.

The quickstart supports provider accounts, a Nous Portal subscription route, and custom endpoints. An external API follows its provider’s rates and quotas. A local endpoint requires compatible hardware and a suitable model. The current quickstart requires a model context window of at least 64,000 tokens, so not every local model configuration qualifies.

Check model and tool billing before enabling recurring tasks. Stored memory and generated skills also require review. Their presence does not guarantee accurate recall or correct automation.

Checked against official documentation on October 23, 2026. We did not install Hermes or test a live automation for this listing.

Verify memory, skills and scheduled work separately

Hermes uses bounded MEMORY.md and USER.md stores under the active profile’s memories/ directory. They are loaded as a snapshot at the start of a session. A promise to remember is not evidence of a saved entry: inspect the file and open a fresh session. Profiles have separate state; do not point two agent processes at the same Hermes home. Memory documentation.

Skills are reusable instructions loaded when relevant, not model training or a guarantee of improvement. Review the actual SKILL.md before recurring use. Scheduled agent tasks require a running gateway, valid model access, an available toolset and a configured delivery target. Check execution and delivery separately. Skills, scheduled tasks.

Our Hermes setup guide walks through a memory check, a small skill and a paused scheduling trial. For hosted and desktop alternatives, see the personal agent comparison.

Quick Info

Hermes Agent icon
Category
Personal agents
Published on
Documentation reviewed
Pricing
Software MIT-licensed self-hosted software
Models Provider subscription or API charges; local hardware costs apply
Connected tools Connected services and always-on hosting may cost extra