What Is Hermes Agent by Nous Research in 2026?
Hermes Agent is an open-source autonomous agent platform from Nous Research, released under the MIT License. Unlike IDE-bound coding assistants or thin chatbot wrappers, it runs independently on your own server and operates across Telegram, Discord, Slack, WhatsApp, Signal, Email and the command line. Nous Research describes it as "the agent that grows with you," because it keeps persistent memory, writes its own skills from experience, and builds a model of how you work across sessions.
That self-hosted, model-agnostic design is what makes Hermes Agent interesting for marketing and operations teams in 2026. You own the server, you choose the model, and you decide how much access the agent has. It is less a product you log into and more an autonomous teammate you configure and supervise.
Why Run a Self-Hosted Agent in 2026?
Running a self-hosted agent like Hermes in 2026 gives you three things hosted assistants cannot: control over your data, freedom to switch models, and the ability to grant real system access safely through sandboxing. For agencies and D2C teams handling client data, that combination matters. You decide where conversations are stored, which provider sees them, and what the agent is allowed to touch.
- Data ownership: memory, sessions and logs live in your own
~/.hermes/directory, not a vendor cloud - Model freedom: point it at Nous Portal, OpenRouter, OpenAI, NVIDIA NIM or your own endpoint and switch any time
- Real capability: web search, browser automation, vision, image generation and text-to-speech, all under your control
- Always-on automation: natural-language cron scheduling for reports, backups and daily briefings
What Do You Need Before Installing Hermes Agent?
Before installing Hermes Agent in 2026 you need a machine running macOS, Linux, WSL2 on Windows, or Termux on Android, plus an API key from at least one model provider. The installer handles the rest, pulling in Python, Node.js, ripgrep and ffmpeg automatically. The one hard requirement on the model side is a context window of at least 64,000 tokens, since smaller windows cannot sustain the multi-step tool-calling the agent relies on.
How Do You Install Hermes Agent in 2026?
Installing Hermes Agent in 2026 takes a single command on most systems. The installer clones the repo, creates a virtual environment, installs all dependencies, and registers a global hermes command. Pick the line that matches your platform, then reload your shell.
macOS, Linux, WSL2 or Termux
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
source ~/.bashrc # or ~/.zshrcWindows (PowerShell)
iex (irm https://hermes-agent.nousresearch.com/install.ps1)On macOS and Windows you can also use the desktop installer at hermes-agent.nousresearch.com/desktop. To install from source instead, clone github.com/NousResearch/hermes-agent and run ./setup-hermes.sh.
How Do You Set Up a Model and Provider?
Setting up a model in Hermes Agent in 2026 is done through the interactive wizard. Run hermes setup for the full guided flow, or hermes model to jump straight to provider and model selection. Secrets such as API keys are stored in ~/.hermes/.env, while everything else lives in ~/.hermes/config.yaml.
hermes setup # full guided wizard
hermes model # choose provider + model
hermes setup --portal # quick path for Nous Portal subscribers
hermes config set model anthropic/claude-opus-4.6
hermes config set OPENROUTER_API_KEY sk-or-...Hermes Agent in 2026 supports a wide range of providers, so you are not locked to one vendor. The choices below are all first-class options.
| Provider | Notes (2026) |
|---|---|
| Nous Portal | Native option, fastest setup with --portal |
| OpenRouter | Access to 200+ models behind one key |
| OpenAI | Direct GPT-class models |
| NVIDIA NIM, z.ai/GLM, Kimi, MiniMax | Alternative frontier and open-weight options |
| Hugging Face / custom endpoint | Bring your own hosted model |
Whatever provider you choose in 2026, the model must support at least 64,000 tokens of context. Hermes Agent uses that window as working memory for multi-step tool calling, so a smaller context window will cause the agent to lose track mid-task. If replies come back empty, run hermes model to confirm both the model choice and the authentication.
How Do You Start Chatting With Hermes Agent?
Once a model is configured, you start Hermes Agent in 2026 by typing hermes for the classic CLI or hermes --tui for the modern terminal interface, which Nous Research recommends. Inside a session, type / to see all slash commands, and use hermes --continue (or -c) to resume your last conversation.
hermes # classic CLI
hermes --tui # modern terminal UI (recommended)
hermes --continue # resume the previous sessionUseful in-session slash commands include /help (all commands), /tools (list available tools), /model (switch models on the fly), and /save (save the conversation). For multi-line input use Alt+Enter, Ctrl+J or Shift+Enter, and to interrupt a running agent, type a new message and press Enter or hit Ctrl+C.
How Do You Connect Telegram, Discord or Slack?
Connecting a messaging platform to Hermes Agent in 2026 is handled by the gateway. Run hermes gateway setup and follow the prompts to link Telegram, Discord, Slack, WhatsApp, Signal, Email, Home Assistant or Microsoft Teams. This is what turns Hermes from a terminal tool into an always-on agent your team can message like a colleague.
hermes gateway setup # connect a messaging platform
hermes gateway status # check connection healthIf a platform stops responding, re-running hermes gateway setup is the standard fix. In group chats, each sender gets their own isolated session by default, so the agent keeps separate context per person rather than blending everyone into one thread.
How Do You Run Hermes Agent Safely With Sandboxing?
Hermes Agent runs commands through a configurable execution backend, and choosing the right one is the most important safety decision in 2026. By default it can run on your local machine, but for any automated or untrusted workload you should switch to an isolated backend so the agent operates inside a container rather than directly on the host.
| Backend | Where commands run | Best for |
|---|---|---|
| local | Your machine directly | Development, personal use |
| docker | Persistent Docker container | Safe sandboxing, automation |
| ssh | Remote server via SSH | Remote or more powerful hardware |
| modal | Modal cloud sandbox | Ephemeral cloud compute |
| daytona | Daytona workspace | Managed cloud environments |
| singularity | Singularity/Apptainer container | HPC clusters |
Switching to Docker isolation is a one-line change, after which the agent runs every command inside a hardened, persistent container:
hermes config set terminal.backend dockerHow Does Hermes Agent Learn and Use Skills?
Hermes Agent grows over time in 2026 through two mechanisms: persistent memory and auto-generated skills. It stores memory files and a user profile under ~/.hermes/, searches its own past conversations, and writes reusable skills from tasks it has completed. You can also browse and install community skills.
hermes skills browse # see available skills
hermes skills search kubernetes # search by keyword
hermes skills install openai/skills/k8s # install a specific skillInstalled skills are invoked with a slash command, for example /k8s deploy the staging manifest. Memory behavior is configurable in config.yaml under the memory block, where you can toggle memory on or off and set character limits for both stored memory and the user profile.
How Do You Schedule Automated Tasks?
Hermes Agent supports natural-language cron scheduling in 2026, which is one of its strongest features for marketing and operations. You can ask it to run recurring jobs such as a Monday morning performance briefing, a nightly backup, or a weekly competitor scan, and it stores those jobs in ~/.hermes/cron/. This is what makes Hermes useful as an unattended teammate rather than a tool you have to sit in front of.
For a global D2C client in 2026, the highest-value Hermes setup we have seen is not chat at all, it is scheduling. A nightly job pulls the day's ad and analytics numbers, summarizes them against targets, and posts a short briefing to a Slack channel before the team logs in. Pair that with the Docker backend so the data-pulling scripts run in an isolated container, and you have a reliable, auditable automation that costs only model tokens to operate.
Where Does Hermes Agent Store Configuration?
Hermes Agent keeps all its state in the ~/.hermes/ directory in 2026, which makes it easy to back up, inspect or move to another server. Configuration follows a clear precedence: CLI arguments override config.yaml, which overrides .env, which overrides built-in defaults. The rule of thumb is simple: secrets go in .env, everything else goes in config.yaml.
| Path | Holds |
|---|---|
config.yaml | Model, terminal, TTS, compression settings |
.env | API keys and secrets |
SOUL.md | Primary agent identity |
memories/ | Persistent memory files |
skills/ | Agent-created skills |
cron/ | Scheduled jobs |
sessions/ and logs/ | Gateway sessions and logs |
Core configuration commands let you view, edit and validate settings without hand-editing files:
hermes config # view current configuration
hermes config edit # open config.yaml in your editor
hermes config set KEY VAL # set a single value
hermes config migrate # add any missing options interactivelyHow Do You Troubleshoot Hermes Agent?
Troubleshooting Hermes Agent in 2026 follows a short recovery sequence that resolves most issues. Start with hermes doctor to diagnose, then work down the list until the agent responds again.
hermes doctordiagnoses common problemshermes modelreselects the provider and confirms auth (fixes empty replies)hermes setupre-runs the full wizardhermes gateway statusthenhermes gateway setupfixes messaging issueshermes --continueresumes a saved session if context was lost
Common Mistakes to Avoid With Hermes Agent in 2026
- Running automation on the local backend: use Docker or a cloud sandbox for any unattended or untrusted workload
- Choosing a small-context model: anything under 64,000 tokens will break multi-step tasks
- Putting secrets in config.yaml: API keys belong in
.env, never in the shared config file - Skipping the gateway status check: when a bot goes quiet,
hermes gateway statustells you whether it is the platform or the model - No human review on scheduled output: treat unattended briefings and reports as drafts until you have audited a few cycles
What Hermes Agent Means for Marketing and Ops Teams
For marketing and operations teams in 2026, Hermes Agent turns repetitive, cross-platform work into something an agent can own. A self-hosted, model-agnostic agent that schedules its own jobs, learns your workflows, and runs safely in a sandbox is closer to a junior team member than a chatbot. The teams that win with it are the ones who treat setup seriously: the right model, the right backend, and clear supervision of what the agent is allowed to do.