Hermes Studio
The only Hermes web UI with a built-in cron job manager — schedule, monitor, and control autonomous agent tasks without touching a terminal.
Not a chat wrapper. A complete studio — orchestrate multi-agent crews, approve actions, browse memory with a visual knowledge graph, manage skills, and schedule recurring tasks, all from one interface. Built for power users running Hermes Agent locally.
What is Hermes Studio?
Hermes Studio is an open-source, self-hosted web dashboard for Hermes Agent by NousResearch. It turns the Hermes AI agent into a full workspace you can control from your browser — with multi-agent orchestration, cron job scheduling, execution approvals, MCP server management, and 30+ features no other Hermes UI offers. Works with Ollama, OpenAI, Anthropic, and any OpenAI-compatible backend. Built with React, TypeScript, and TanStack. MIT licensed.
- What is Hermes Studio?
- Features
- Cron Job Manager
- What's different from hermes-workspace
- Screenshots
- Quick Start
- Local Models (Ollama, LM Studio, vLLM)
- Docker Quickstart
- Install as App (PWA)
- Mobile Access via Tailscale
- Feature Details
- Troubleshooting
- Uninstall
- Roadmap
- Contributing
- License
✨ Features
- 🤖 Hermes Agent Integration — Direct gateway connection with real-time SSE streaming
- 👥 Multi-Agent Crews — Create named crews of specialised agents, dispatch tasks to all or specific members, watch live activity feeds
- 🗂️ Profile-Scoped Workspaces — Each agent crew member gets an isolated file system view via per-profile workspace roots
- 🕸️ Interactive Knowledge Graph — Force-directed visual graph of your memory's wiki-link relationships with zoom, pan, node drag, and hover highlights
- 🎨 8-Theme System — Official, Classic, Slate, Mono — each with light and dark variants
- 🔒 Security Hardened — Auth middleware on all API routes, CSP headers, path traversal guards, exec approval prompts
- 📱 Mobile-First PWA — Full feature parity on any device via Tailscale
- ⚡ Live SSE Streaming — Real-time agent output with tool call rendering
- 🧠 Memory & Skills — Browse, search, and edit agent memory; explore 2,000+ skills
- ✅ Execution Approvals — Approve, deny, or always-allow agent shell commands from the UI; resolved receipts shown inline
- 📦 Skill Installation — Install/uninstall/toggle skills directly from the browser
- 🐦 X/Twitter Plugin Example — Use Hermes Tweet as a native Hermes Agent plugin for X/Twitter search, read, and guarded action workflows:
hermes plugins install Xquik-dev/hermes-tweet --enable - ⏰ Cron Job Manager — The only agent UI with a full scheduler: create, edit, pause, trigger, and monitor jobs; manual triggers stream live tool events via SSE directly into the job card
- 🔐 Permissions & Toolsets — Configure approvals, command allowlist, toolsets, security scanner, code limits, and reasoning from Settings UI
- 💾 Session Persistence — Auth tokens, sessions, and active runs survive server restarts via Redis (auto-connects, graceful fallback)
- 🔀 Visual Workflow Builder — Build and run DAG-structured task pipelines for your crews; tasks run in topological order with live per-node status
- 📋 Crew & Conductor Templates — 7 built-in crew templates + 4 conductor templates (Research, Build, Review, Deploy) plus save and manage your own; unified template system with
templateTypefield - 💰 Cost Tracking — per-crew token usage (input/output) and estimated API cost per agent; Usage tab on every crew with model-aware price table and reset control
- 🔌 MCP Server Management — add, edit, and remove MCP servers from the Settings UI; saves directly to
~/.hermes/config.yamland triggers a live reload; no manual file editing required - 🧬 Agent Library — create, edit, and delete custom agents with bespoke system prompts, emoji, role labels, and model overrides; built-in agents include pre-written system prompts you can copy and customize; custom agents appear in the crew builder and template picker alongside built-in personas
- 🕵️ Audit Trail — chronological timeline of every tool call, user message, and approval request across all sessions; filter by session, event type, and date range; expand tool events to inspect full args and results
- 📋 Clone Crew — duplicate any existing crew in one click; mints fresh sessions for every member and opens the new crew immediately; available on the crew card (hover) and from the crew detail header
- 🧪 Test Suite — vitest unit tests for core stores and utilities (199 tests across 17 files, all passing); Playwright e2e smoke tests for all major pages; GitHub Actions CI with visible status badge
- ⚡ Fast Mode — one-click toggle in the chat composer activates Hermes v0.9.0's priority queue for OpenAI and Anthropic models; persists per-session
- 📋 Logs Viewer —
/logsscreen pulls the last 500 lines from~/.hermes/logs/(Hermes v0.8.0 centralised logging); filter All/Errors, live search, color-coded by level, auto-scroll - 💾 Backup & Import — one-click Hermes data backup and restore from Settings (Hermes v0.9.0)
- 📡 iMessage, WeChat & WeCom — BlueBubbles (iMessage), WeChat, and WeCom/Enterprise added to the platform integration settings alongside Telegram, Discord, Slack, and Signal (Hermes v0.9.0)
- 🎯 Conductor V2 — Gateway-native mission orchestration with animated SVG office (3 layouts: Grid, Roundtable, War Room), pixel-art agent avatars, real Hermes gateway integration, live session polling, worker monitoring, settings drawer (model selection, max parallel, supervised mode), mission history with cost tracking, quick actions (Research/Build/Review/Deploy), abort/pause/retry controls
- 📊 Operations Dashboard — Unified view of all running agents across crews and conductor missions; grid and outputs toggle; status filters; per-crew Operations tab also available on crew detail screens
- 📋 Tasks / Kanban Board — Five-column Kanban (Backlog → Todo → In Progress → Review → Done) with native HTML5 drag-and-drop; create/edit tasks with priority, tags, and source links; conductor missions auto-create cross-linked tasks
- 📊 Rate Limit Display — provider usage meter now surfaces API rate-limit headers (requests/tokens remaining, reset countdown) captured by Hermes v0.9.0 from LLM provider responses
- 🖥️ System Health Panel — fixed footer bar showing live CPU %, memory, disk usage, and uptime; color-coded thresholds (green/amber/red); toggle in Settings → Display
- 📈 Token Usage Chart — 14-day area chart in the usage modal breaks down daily input vs output tokens across all sessions
- 🔬 Event Analytics —
/analyticsscreen with aggregate event stats, daily volume bar chart, and top-15 tool frequency chart sourced directly from the SQLite event store - 🪪 Identity File Editor — Settings → Identity lets you read and write
SOUL.md,persona.md, andCLAUDE.mddirectly from the browser; no terminal required - 🧩 Patterns & Corrections Viewer —
/patternsscreen to browse agent-learned patterns and manage user corrections stored inMEMORY.md - 🕐 Session History Archive — two-pane session browser at
/session-history; sortable by date/model/tokens/cost; lazy-loads full message thread per session - ⚙️ Systemd Auto-start — Settings → Auto-start generates and installs a systemd user service unit; install/start/stop/enable/disable/uninstall from the UI with live status display
⏰ Cron Job Manager — a feature no other UI has
Every other Hermes/Claude web interface treats the agent as a request-response tool. You send a message, you get a reply.
Hermes Studio is the only one that lets you schedule the agent as a background worker — running prompts on a timer, automatically, while you do something else.
No crontab -e. No shell scripts. No babysitting.
From the Jobs tab you can:
- Create jobs with a natural-language prompt and a schedule — presets (every 15 min, hourly, daily, weekly) or any custom cron expression
- Pick delivery channels — route job output to T