Nous Research Ships Hermes Agent Profile Builder for Guided Web-based Configuration

🔄 Update — 12. June 2026: Community Project Introduces Native Windows Desktop App for Hermes Agent

An independent developer has built a native Windows desktop app around the Hermes Agent framework. This version bypasses the standard web interface in favor of a fast, locally running WinUI 3 application.

What’s new?

WinUI 3 & .NET 10 Interface: A completely native Windows interface replacing the Node/Web-view stack, offering a fast and lightweight local experience.
Soul System & 94 Skills: Swap customized “soul templates” for different personas and access 94 built-in tools spanning files, browser automation, and code sandboxes.
Mistakes & Habits Journals: Active memory features that automatically extract past errors to prevent repeating them and reinforce successful interaction patterns.

Why this adds to the article

This community project demonstrates how rapidly developers are building user-friendly desktop alternatives with advanced local memory management on top of Hermes Agent. It reinforces the original article’s focus on the shift toward modular, visually accessible local AI agent infrastructures.

Summary

Nous Research has released a browser-based Profile Builder for Hermes Agent, simplifying the creation and configuration of isolated agent profiles. Instead of executing complex command-line instructions, developers can now configure identity, LLM providers, built-in or catalog skills, and Model Context Protocol (MCP) servers through a guided web dashboard. The Profile Builder creates isolated agent instances on the same machine, preventing state or data collisions.

What happened?

By introducing the Profile Builder—launched via hermes dashboard at http://127.0.0.1:9119—Nous Research offers a visual alternative to the previously CLI-only workflows of Hermes Agent. Running locally, the dashboard collects configuration details and writes them directly to the agent’s profile files. Developers can configure five settings groups in a single flow:

Identity: The profile name and description. The chosen name becomes a shell command alias (e.g., coder chat starts the “coder” profile). Deeper agent personality is defined in SOUL.md.
Model & Provider: Support for endpoints like Nous Portal, OpenRouter, NVIDIA, OpenAI, or custom OpenAI-compatible APIs.
Built-in Skills: Toggle native tools on or off for each profile.
Skills Hub: Install external capabilities from catalogs by identifier.
MCP Servers: Connect tool servers via command line (stdio) or HTTP URLs.

Why it matters

Managing multiple specialized agents on a single machine was previously cumbersome and carried a risk of state collisions. The Profile Builder makes creating isolated environments seamless. Each profile lives in its own home directory with a dedicated config.yaml, .env, SOUL.md, separate memory, sessions, and state database. Furthermore, Hermes Agent demonstrates an excellent practice for context-saving: it initially reads only brief descriptions of skills, loading the full SKILL.md content on demand only when a task requires it.

Evidence

Official documentation and technical repositories from Nous Research support these new capabilities:

Dashboard Startup: The local server is initialized with hermes dashboard and defaults to port 9119 on loopback.
Security Protocols: Non-loopback binds fail closed unless an authentication provider is configured.
Package Dependency: The dashboard component requires installing the web extras package via pip install 'hermes-agent[web]'.

Analysis

Nous Research’s move highlights the trend toward modularity and visual accessibility in AI agent tooling. While advanced users can still use the CLI, the web interface dramatically lowers the barrier to entry. Strict profile isolation avoids accidental credential leaks or conflicting memory states. Storing profiles in dedicated folders reflects clean software architecture. Standardized Model Context Protocol (MCP) integration also makes it straightforward to grant agents secure access to local filesystems or remote APIs.

Practical Takeaways

Developers can spin up dedicated profiles for distinct use cases:

Focused Coding Assistant: A profile with a code-aware model, filesystem MCP, and Git skills, with its memory scoped strictly to one project.
Research Agent: A profile equipped with web-search and extraction skills, keeping its findings isolated from other agents.
Operations Bot: A gateway-driven bot with dedicated tokens and scheduled cron reports. To install the dashboard, run:

pip install 'hermes-agent[web]'

Open Questions

How quickly will the developer community adopt the visual Profile Builder over the familiar CLI commands?
What security vulnerabilities might emerge if developers misconfigure the dashboard to bind to public IP addresses without proper authentication?