--- title: "Nous Research's Hermes Agent Now Runs on Windows Without WSL or Docker" subtitle: "Nous Research's self-improving AI agent lands on Windows natively, no WSL required, as NVIDIA backs it at GTC Taipei" url: "https://alphasignal.ai/news/nous-research-s-hermes-agent-now-runs-on-windows-without-wsl-or-docker" author: "AlphaSignal Newsroom" author_url: "https://alphasignal.ai/editorial-team" company: "Nous Research" company_url: "https://alphasignal.ai/company/nousresearch" published: "2026-06-01T00:01:22.000Z" updated: "2026-06-16T20:42:39.516Z" type: "News" topics: - "Agents" - "Development" subtopics: - "Agent Frameworks" - "Computer Use" - "Tool Use" upvotes: 5349 read_time: "6 min" source_url: "https://hermes-agent.nousresearch.com/docs/user-guide/windows-native" comments: 0 --- # Nous Research's Hermes Agent Now Runs on Windows Without WSL or Docker > Nous Research's self-improving AI agent lands on Windows natively, no WSL required, as NVIDIA backs it at GTC Taipei [Hermes Agent](https://hermes-agent.nousresearch.com/docs/user-guide/windows-native), the open-source self-improving AI agent from Nous Research, now runs natively on Windows 10 and Windows 11 , no WSL, no Docker, no Cygwin required. The release, announced alongside the Hermes Desktop public preview at v0.15.2, caps a rapid expansion that has taken the project from a Linux-only CLI tool to a full cross-platform agent with a GUI in just a few months. ## The agent that refuses to forget Hermes Agent is an open-source autonomous AI agent built by Nous Research, first released in February 2026. Unlike traditional one-shot agents, Hermes is the only open-source agent with a true built-in learning loop , it doesn't just complete tasks, it remembers, builds reusable skills, and gets smarter every single time you use it. All data stays on your machine, with no telemetry, no tracking, and no cloud lock-in. It connects to over 200 AI models across 20 messaging platforms, learns from every task through a three-layer memory system, and generates reusable skills that make it measurably better at recurring workflows over time. The self-improvement mechanism is the headline differentiator: the agent writes a new reusable skill on its own after completing a complex task of five or more tool calls, with zero human authoring required. ## What "native Windows" actually means Before this release, Windows users had two options: run Hermes through WSL2 (the Linux compatibility layer built into Windows), or skip it entirely. Native support changes that equation significantly. Hermes runs natively on Windows 10 and Windows 11 with no WSL, no Cygwin, and no Docker. The install is a single PowerShell one-liner: ``` iex (irm https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.ps1) ``` No admin rights are required. The installer goes to `%LOCALAPPDATA%\hermes\` and adds `hermes` to your User PATH , open a new terminal after it finishes. For those who prefer a double-click experience, a GUI installer is also available that provisions Python, Node, PortableGit, and all dependencies automatically. Under the hood, the installer does a lot of heavy lifting so you don't have to: - Bootstraps `uv`, Astral's fast Python manager, and installs Python 3.11 , no existing Python needed - Installs Node.js 22 for the browser tool and WhatsApp bridge - Downloads a portable Git (~45 MB) if none is detected, with no admin rights or registry changes - Runs a tiered dependency install that falls back gracefully if GitHub rate-limits a fetch - Runs `hermes setup` to walk you through model and provider selection ## Almost everything works , with one exception The feature parity between native Windows and WSL2 is remarkably close. Here's what runs natively: - CLI (`hermes chat`, `hermes setup`, `hermes gateway`) - Interactive TUI (`hermes --tui`) - Messaging gateway across Telegram, Discord, Slack, WhatsApp, and 15+ other platforms - Cron scheduler and auto-start at login via Scheduled Tasks - Browser tool (Chromium via Node) - MCP servers (stdio and HTTP) - Local model support via Ollama, LM Studio, or llama-server - Full web dashboard (sessions, jobs, metrics, config) The one gap: the dashboard's `/chat` tab embeds a real terminal via a POSIX PTY. Native Windows has no equivalent primitive; treat it as future work. The rest of the dashboard works natively , only that one tab shows a "use WSL2 for this" banner. ## The engineering work behind the port Getting a POSIX-first agent to run cleanly on Windows required fixing a surprisingly long tail of platform-specific bugs. A long tail of native-Windows fixes shipped alongside the beta , taskkill-based subprocess management, MinGit auto-install, Microsoft Store Python stub detection, npm prefix handling, native PTY paths, signal handling differences, foreground process management, ANSI sequence handling, path normalization, file-locking semantics, and many more. One particularly subtle bug: Python's `os.kill(pid, 0)` , normally a harmless "is this process alive?" check on Linux , maps to `CTRL_C_EVENT` on Windows at the integer level. The result was that any code checking whether the gateway was running was silently killing it instead. The team migrated all 14 affected call sites across 11 files to use `psutil.pid_exists()` and added a CI check to catch any future regressions. Shell command execution uses Git Bash as the bridge, the same strategy Claude Code uses. This sidesteps the POSIX-vs-Windows gap without rewriting every tool. The installer sets `HERMES_GIT_BASH_PATH` explicitly so fresh PowerShell sessions don't need to rediscover it. ## The NVIDIA connection The Windows launch didn't happen in a vacuum. At GTC Taipei, NVIDIA unveiled RTX Spark, a new superchip that reinvents Windows PCs for the era of personal AI agents. NVIDIA cited open-source projects like Hermes Agent achieving record-breaking numbers on developer networks like GitHub and OpenRouter, while noting that broad adoption has been limited by the inability to run agents securely and privately on users' primary PCs , a gap NVIDIA and Microsoft are partnering to address. The NemoClaw installer , NVIDIA's stack for running agents on-device , now adds support for Hermes Agent alongside OpenClaw. Hermes Desktop was first demoed in Jensen Huang's GTC keynote, giving the project a level of industry visibility that few open-source agent frameworks have achieved this quickly. ## What it's actually good at Like other popular agents, Hermes integrates with messaging apps, can access local files and applications, and runs 24/7. But self-evolving skills set it apart: every time the agent encounters a complex task or receives feedback, it saves its learnings as a skill so it can adapt and improve over time. Hermes treats sub-agents as short-lived, isolated workers dedicated to a sub-task , with a focused context and set of tools. This keeps task organization tidy and allows Hermes to run with smaller context windows, which is ideal for local models. Beyond task automation, Hermes is also a platform for generating training data, running RL experiments, and exporting trajectories for fine-tuning , generating thousands of tool-calling trajectories in parallel with automatic checkpointing. Where it's less polished: some users have noted that Hermes works faster than competitors but can be less stable, and its memory feels "increasingly shaped by accumulated workflows" rather than giving the user explicit control. The Windows installer is also not code-signed, so Windows SmartScreen will flag it on first launch, requiring users to click "More info" then "Run anyway." ## Getting started Hermes Agent is open source under the MIT license. The base agent is free to run with your own API keys. A Nous Portal subscription covers both the model and all tool APIs (browser, image generation, TTS) through a single OAuth login , useful for avoiding the per-tool API key setup that the docs describe as the highest-friction part of a Windows install. For local inference, the desktop app shipped as a public preview at v0.15.2 with native builds for macOS, Windows, and Linux. It includes one-click install, in-app self-update, drag-and-drop chat files, an inline model picker, concurrent profiles, Simplified Chinese UI, and remote gateway sign-in. The [GitHub repo](https://github.com/NousResearch/hermes-agent) and the [official Windows guide](https://hermes-agent.nousresearch.com/docs/user-guide/windows-native) are the best starting points. The broader picture here is that a self-improving, persistent agent , one that genuinely accumulates capability across sessions , is now a one-liner install away on the world's most common desktop OS. Whether that changes how developers think about local agents is the real question worth watching. ## Sources - [X post by @nousresearch](https://x.com/nousresearch/status/2061236625925886252) - [Windows (Native) Guide | Hermes Agent](https://hermes-agent.nousresearch.com/docs/user-guide/windows-native) - [Hermes Unlocks Self-Improving AI Agents, Powered by NVIDIA RTX PCs and DGX Spark | NVIDIA Blog](https://blogs.nvidia.com/blog/rtx-ai-garage-hermes-agent-dgx-spark/) - [NVIDIA and Microsoft Reinvent Windows PCs for the Age of Personal AI | NVIDIA Newsroom](https://nvidianews.nvidia.com/news/nvidia-microsoft-windows-pcs-agents-rtx-spark) - [Hermes Agent Desktop App | Medium](https://medium.com/@tentenco/hermes-agent-desktop-app-everything-you-need-to-know-about-nous-researchs-self-improving-ai-agent-3cb59bd31e5f) - [Hermes Agent Releases | GitHub](https://github.com/NousResearch/hermes-agent/releases)