Nous Research Ships Hermes Agent Profile Builder: Identity, Model, Skills, and MCP Servers in One Dashboard Flow

Nous Research has shipped a Profile Builder for Hermes Agent. It lives inside the project’s local web dashboard. Standing up a distinct agent used to mean several CLI steps. The builder now walks you through one guided flow. In that flow you define an agent’s identity. You pick a model and provider. You choose built-in and optional skills. You install skills from the hub. You attach MCP servers. Hermes Agent is Nous Research’s open-source, self-improving agent. It runs on the CLI, a desktop app, and messaging platforms. Profiles were previously assembled mostly through terminal commands. The Profile Builder brings those pieces into a browser form. Profile Builder A profile in Hermes is a separate home directory. Each profile holds its own config.yaml, .env, and SOUL.md. It also keeps separate memory, sessions, skills, cron jobs, and a state database. Profiles let you run isolated agents on one machine. A coding agent and a research agent never share state. This is the unit the builder produces. You launch the dashboard by running hermes dashboard. It opens at http://127.0.0.1:9119 in your browser. The default bind is loopback, so no data leaves localhost. The builder collects the same inputs the CLI profile commands accept. It then writes them into the profile’s files. The Fields the Builder Configures The builder gathers five groups of settings in one place: First is identity, a name and a description. The name also becomes a shell command alias. Create a profile named coder and you get coder chat. Deeper personality lives in the profile’s SOUL.md file. Second is the model and provider. Hermes supports Nous Portal, OpenRouter, NVIDIA, OpenAI, and more. You can also point at your own OpenAI-compatible endpoint. Third is built-in skills, toggled on or off per profile. Fourth is Skills Hub installs, pulled from external catalogs by identifier. Fifth is MCP servers, added by URL or by local command. Two of these terms deserve a short explanation. Skills are SKILL.md files with a name, a description, and a procedure. The agent reads short descriptions cheaply. It loads a skill’s full content only when a task needs it. So adding many skills does not bloat every request. MCP servers expose external tools through the Model Context Protocol. Hermes accepts HTTP servers via a URL. It also accepts stdio servers via a local command. A Nous-approved catalog offers one-click installs, prompting inline for any keys. GUI Flow vs the CLI Sequence The builder does not replace the CLI. It mirrors it in a form. The table below maps each step to its command equivalent. StepProfile Builder (dashboard)CLI equivalentCreate and nameName fieldhermes profile create coderDescriptionDescription field--description "..." or hermes profile describeModel and providerModel pickercoder config set model <id>Built-in skillsTogglescoder skills list / toggleSkills Hub installSearch and installcoder skills install <slug>MCP serversAdd or catalog installedit mcp_servers / coder mcp install Both paths edit the same profile directory. The builder is the lower-friction entry point. The CLI remains the scriptable one. Use Cases With Examples The first use case is a focused coding assistant. Give it a code-aware model and a filesystem MCP server. Add Git and testing skills. Keep its memory scoped to one project. The second is a research agent. Pair a capable model with web-extract skills. Its findings stay separate from your other agents. Cloning the profile later preserves that separation. The third is an operations bot. Attach a messaging channel and schedule cron reports. Each profile runs its own gateway and bot token. Token locks block two profiles from sharing a token by accident. In each case, the builder produces one isolated agent. You can run several without state collisions. What the Builder Writes The builder edits files the CLI already reads. Model and provider land in the profile’s config.yaml. MCP servers populate the mcp_servers block in that file. API keys go to the profile’s .env. The equivalent CLI sequence for a research profile looks like this: Copy CodeCopiedUse a different Browserhermes profile create researcher \ --description "Reads source code and external docs, writes findings." researcher setup # configure API keys + model researcher config set model anthropic/claude-sonnet-4 researcher skills install openai/skills/k8s A matching config.yaml for that profile. Note that mcp_servers is a map keyed by server name, not a list: Copy CodeCopiedUse a different Browser# ~/.hermes/profiles/researcher/config.yaml model: default: anthropic/claude-sonnet-4 mcp_servers: filesystem: command: npx args: ["-y", "@modelcontextprotocol/server-filesystem", "/home/user/projects"] A remote HTTP MCP server uses url and headers instead of command: Copy CodeCopiedUse a different Browsermcp_servers: docs: url: "https://mcp.example.com/mcp" headers: Authorization: "Bearer ${DOCS_API_KEY}" Reaching the builder requires the dashboard extra. The base install ships without the HTTP stack. Install it with one command: Copy CodeCopiedUse a different Browserpip install 'hermes-agent[web]' Strengths and Limitations Strengths: One flow replaces several CLI steps for a new profile. Identity, model, skills, and MCP sit in one place. Output stays in standard config.yaml and .env files. The dashboard runs locally and binds to loopback by default. The catalog handles MCP and Skills Hub installs inline. Limitations: Profiles do not sandbox filesystem access on the local backend. A non-loopback bind fails closed unless an auth provider is configured. Skill and MCP changes take effect on the next session or gateway restart. The builder surfaces controls that also live on separate dashboard pages. Marktechpost’s Interactive Explainer @import url('https://fonts.googleapis.com/css2?family=Space+Grotesk:wght@500;600;700&family=JetBrains+Mono:wght@400;500;600&display=swap'); #mtp-hpb-700 *{box-sizing:border-box!important;margin:0!important;padding:0!important} #mtp-hpb-700 hr,#mtp-hpb-700 p:empty,#mtp-hpb-700 del,#mtp-hpb-700 s{display:none!important} #mtp-hpb-700{ --bg:#06140f!important; --panel:#0b1f19!important; --line:#1b3d35!important; --line2:#24514a!important; --teal:#2fd6bd!important; --teal2:#13a892!important; --cream:#f1ead6!important; --soft:#c4d0ca!important; --muted:#7e948c!important; --gold:#e7c478!important; --mono:'JetBrains Mono',ui-monospace,Menlo,Consolas,monospace!important; --disp:'Space Grotesk',system-ui,-apple-system,'Segoe UI',Roboto,sans-serif!important; all:initial!important; display:flex!important; flex-direction:column!important; width:100%!important; max-width:700px!important; height:500px!important; margin:0 auto!important; background:var(--bg)!important; color:var(--cream)!important; font-family:var(--disp)!important; border:1px solid var(--line)!important; border-radius:14px!important; overflow:hidden!important; -webkit-font-smoothing:antialiased!important; } /* HEADER */ #mtp-hpb-700 .w-head{flex:0 0 auto!important; padding:14px 18px 12px!important; border-bottom:1px solid var(--line)!important; background:radial-gradient(520px 160px at 88% -30%, rgba(47,214,189,.16), transparent 65%)!important; display:flex!important; align-items:center!important; gap:10px!important} #mtp-hpb-700 .w-dot{width:7px!important;height:7px!important;border-radius:50%!important;background:var(--teal)!important;box-shadow:0 0 9px var(--teal)!important;flex:0 0 auto!important;animation:w7pulse 2.4s ease-in-out infinite!important} @keyframes w7pulse{0%,100%{opacity:1}50%{opacity:.35}} #mtp-hpb-700 .w-ttl{min-width:0!important} #mtp-hpb-700 .w-eyebrow{font-family:var(--mono)!important;font-size:9.5px!important;letter-spacing:.2em!important;color:var(--teal)!important;text-transform:uppercase!important;white-space:nowrap!important} #mtp-hpb-700 .w-h1{font-size:17px!important;font-weight:700!important;color:var(--cream)!important;line-height:1.2!important;white-space:nowrap!important;overflow:hidden!importan