Nous Research has shipped a Profile Builder for its open-source Hermes Agent, replacing multiple command-line steps with a single guided web interface. The builder, accessible through a local dashboard, lets developers define an agent's identity, select a model and provider, choose skills, and attach external tools in one unified flow.

What Problem Does the Profile Builder Solve?

Previously, setting up a distinct Hermes Agent required developers to run several terminal commands to assemble configuration files, environment variables, and skill definitions. The Profile Builder brings all of these pieces into a browser-based form, dramatically reducing friction for new users. Each profile created in Hermes is a separate home directory containing its own configuration, environment settings, memory, sessions, and state database, meaning multiple agents can run on one machine without interfering with each other.

Developers launch the dashboard by running a single command, which opens at localhost on port 9119. Because the default binding is loopback, no data leaves the local machine. The builder collects the same inputs that CLI profile commands accept and writes them directly into the profile's configuration files.

How to Set Up a Hermes Agent Profile Using the Builder

• Define Identity: Enter a profile name and description. The name automatically becomes a shell command alias, so a profile named "coder" generates a "coder chat" command. Deeper personality traits can be customized in the profile's SOUL.md file.
• Select Model and Provider: Choose from supported providers including Nous Portal, OpenRouter, NVIDIA, OpenAI, or point to a custom OpenAI-compatible endpoint. The model can be switched at any time after initial setup.
• Configure Skills and Tools: Toggle built-in skills on or off per profile, install skills from an external hub by identifier, and attach Model Context Protocol (MCP) servers via URL or local command. Skills are loaded only when needed, so adding many does not bloat every request.

The builder mirrors the CLI's functionality in a graphical form. Both paths edit the same underlying profile directory, so developers can switch between the dashboard and terminal commands without conflicts. The builder serves as the lower-friction entry point for new users, while the CLI remains the scriptable option for automation.

What Makes Isolated Agent Profiles Valuable?

The Profile Builder enables three primary use cases that demonstrate the value of isolated agents. First, a focused coding assistant can be configured with a code-aware model, a filesystem MCP server, Git and testing skills, and memory scoped to a single project. Second, a research agent can pair a capable model with web-extraction skills, keeping findings separate from other agents. Third, an operations bot can attach to a messaging channel and schedule automated reports, with each profile running its own gateway and bot token to prevent accidental token sharing.

This isolation prevents state collisions. A coding agent and a research agent never share memory, sessions, or configuration, allowing developers to run specialized assistants tailored to specific workflows without cross-contamination. The builder produces one isolated agent per profile, and multiple profiles can run simultaneously on the same machine.

Technical Architecture and Configuration

The builder edits standard configuration files that the CLI already reads. Model and provider settings land in the profile's config.yaml file. MCP servers populate the mcp_servers block as a map keyed by server name. API keys are stored in the profile's.env file. Remote HTTP MCP servers use URL and headers, while local servers use command and arguments.

The dashboard requires the web extra to be installed. The base Hermes Agent installation ships without the HTTP stack, so developers must run "pip install 'hermes-agent[web]'" to access the Profile Builder. The builder's strengths include consolidating multiple CLI steps into one flow, keeping all settings in one place, running locally with loopback binding by default, and handling catalog installs inline. Limitations include the lack of filesystem sandboxing on the local backend, the requirement for an auth provider when binding to non-loopback addresses, and the need to restart sessions or gateways for skill and MCP changes to take effect.

Hermes Agent itself is Nous Research's open-source, self-improving agent that runs on the CLI, a desktop app, and messaging platforms. The Profile Builder represents a significant usability improvement, lowering the barrier to entry for developers who want to deploy multiple specialized AI assistants without deep terminal expertise. By combining identity definition, model selection, skill management, and tool integration into a single dashboard, the builder makes agent deployment more accessible while maintaining the flexibility and isolation that power users require.