Favicon of Hermes HUD

Hermes HUD

A local AI agent dashboard for Hermes on macOS, Linux and WSL, with live chat, memory inspection, cost analytics and redacted session replays. MIT licensed.

Hermes HUD is a local browser dashboard for people running the Hermes AI agent who want to inspect its memory, follow its work and understand its costs. It runs on macOS, Linux and WSL and requires an existing Hermes agent. The project is open source under the MIT license.

The dashboard brings agent health, spending, model usage and provider problems into one overview. Separate views cover persistent memory, skills, session history, scheduled jobs and projects. Live updates let you follow activity without refreshing the page, and built-in chat lets you talk to the agent from the same interface.

For troubleshooting, it provides health diagnostics and token and cost analytics for each model. It also shows how web search, image generation, speech and browser automation connect through Nous Tool Gateway or direct keys. Plugin controls expose runtime status and let you enable, disable or update agent and dashboard extensions.

Session replays provide timelines and exportable records for issues, pull requests or sharing. Exports stay local by default. Safe Share Mode removes sensitive content such as tool arguments, terminal output, reasoning, credentials and local paths. Export formats include JSON, Markdown, standalone HTML and PNG share cards.

The server accepts local connections by default. Optional publishing sends sanitized replays to a Git repository for a static gallery, including GitHub Pages, only when you trigger a sync. The web interface works independently of the hermes-hud terminal companion, and both can read the same agent data.

Similar to Hermes HUD