Stack
Hermes Agent is a free, open-source AI agent made by Nous Research. We build with it to give marketing teams customer-facing conversational agents on WhatsApp or Telegram and scheduled reports delivered to a team chat. It runs on any model and on infrastructure the team controls, which matters for teams with data privacy or model-choice requirements.
Book a callWhat it is
Hermes Agent is a self-improving, open-source AI agent. It creates skills from experience, improves them during use and keeps persistent memory of past conversations. It runs from one gateway on Telegram, Discord, Slack, WhatsApp, Signal, email and CLI. A desktop app is available for macOS, Windows and Linux. Hermes Agent works with any model: Nous Portal, OpenRouter, OpenAI or a private endpoint. It includes a cron scheduler for unattended reports and audits, isolated subagents for parallel work, web search, browser automation, vision, image generation and text-to-speech. It runs locally, on Docker, SSH or serverless backends such as Modal or Daytona, or on a VPS. Nous Portal is an optional subscription for model and tool credits.
Maker: Nous Research · Status: Available, free and open source under MIT license · Official site: hermes-agent.nousresearch.com/
The marketing jobs we build with it
A conversational agent on WhatsApp or Telegram answers questions from prospects and qualifies them against a set of criteria. When a prospect meets the threshold, the agent routes them to the sales team or books a call.
Hermes Agent's built-in cron scheduler runs unattended audits and reports on a set cadence. We use it to deliver a weekly performance summary to a Slack or Discord channel without anyone pulling the data manually.
Teams with data privacy requirements or a preference for a specific model can deploy Hermes Agent on a VPS, Docker or a serverless backend. The agent connects to any model endpoint, so the team can switch providers.
Hermes Agent keeps persistent memory across sessions and builds skills from experience. A research agent built on it recalls what it learned in earlier sessions and refines its approach each time it runs.
When we choose it, and when we don't
Honest limits
Works well with
Related reading
FAQ
Hermes Agent is an open-source AI agent made by Nous Research. It runs on any model, connects to messaging channels including WhatsApp, Telegram, Slack and Discord, and keeps persistent memory across sessions. It is free under the MIT license. Model and hosting costs are separate.
Hermes Agent runs on Telegram, Discord, Slack, WhatsApp, Signal, email and CLI from one gateway. A desktop app is available for macOS, Windows and Linux. One agent can handle multiple channels without separate deployments.
Yes. Hermes Agent creates skills from experience and improves them during use. For customer-facing workflows, those skills should be reviewed before they go into production, because the agent changes its own behavior over time.
Our builds most often start with a WhatsApp or Telegram qualification agent or a scheduled report delivered to a team chat. The agent connects to a model of the team's choice and runs on infrastructure the team controls. The persistent memory and skill-building loop is useful for research agents that need to carry context across weeks of work.
A developer or technical team member sets up and maintains the host. Hermes Agent runs on a local machine, a VPS, Docker or a serverless backend such as Modal or Daytona. Teams that want a managed service with no hosting responsibility would be better served by Claude agents or Hyperagent.
Tell us the repetitive work. We will tell you whether Hermes Agent is the right place to build it.
Book a call
