// the find
EKKOLearnAI/hermes-studio
Web dashboard for Hermes Agent — multi-platform AI chat, session management, scheduled jobs, usage analytics
Hermes Studio is a web dashboard and desktop app for Hermes Agent, a self-hosted AI agent runtime. It wraps the agent with session management, multi-platform channel integrations (Telegram, Discord, Slack, WeChat, etc.), scheduled jobs, and a file browser. Aimed at developers who want a local-first, self-hosted alternative to managed AI chat products.
The BFF-plus-bridge architecture is well thought out — the Koa server proxies to a Python agent bridge over ZMQ, which keeps the Node layer thin and lets the actual agent logic stay in Python. Multi-profile support with per-profile credential isolation is genuinely useful for running separate agent identities without credential bleed. The distribution story (npm CLI, Docker, Electron desktop, all from one codebase) is solid for a self-hosted tool. The channel integration breadth — including WeChat QR login — covers the real-world mix of platforms that enterprise and power users actually care about.
The BSL-1.1 license is a meaningful constraint: you can self-host for personal use, but commercial deployment or building a product on top requires a commercial agreement with EKKOLearnAI. That's a trap for anyone who builds internal tooling and then scales it. The default credentials ship as admin/123456 with a post-login prompt to change them — fine for desktop, but anyone who docker-composes this onto a VPS and forgets that step is exposed. The architecture docs mention the frontend is designed for 'multi-agent extensibility' with Hermes namespacing, but there's no actual plugin contract or stable API surface documented for third-party agents — it's aspirational architecture. Tight coupling to the Hermes Agent runtime (the bridge requires `run_agent.py` from a specific source layout) means switching or extending the underlying agent is nontrivial.