freejarvis: app and UI for Hermes Agent

Voice command deck that runs on its own model or acts as the face of a self-hosted Hermes Agent

What freejarvis does

The deck has two brains and one face. Both emit the same event types, such as assistant.delta, tool.started and approval.required, so panels, captions and the animated orb behave the same either way. Setting HERMES_URL attaches Hermes, whose agent loop, memory, skills and run_approval gating are then used, and Hermes cron jobs can deliver to Telegram or Discord. Without it, freejarvis uses its own loop, eight built-in tools, a local SQLite session store and an in-process scheduler.

Speech recognition is the browser's, so in Chrome and Edge the audio goes to Google while the talk key is held; Safari recognizes on-device and Firefox falls back to typing. Text-to-speech is local unless you add an ELEVENLABS_API_KEY. The README is open about limits: no wake word, no login because the deck binds to localhost, memory is a list with substring search rather than embeddings, and the scheduler stops when the container is down. Hermes panels read /v1/toolsets, /v1/skills and /api/jobs, and show a notice if your Hermes build does not serve one.

Key features

Hold-to-talk voice input with spoken replies and a live caption line
Attaches to a self-hosted Hermes Agent through HERMES_URL
Live panels for Hermes sessions, cron jobs, skills, toolsets and approvals
Own agent loop with eight built-in tools for any OpenAI-compatible model
Local text-to-speech with optional ElevenLabs voice
Docker Compose setup that serves the deck at http://localhost:4333

When to use it

Talking to a self-hosted Hermes Agent by voice from a browser
Watching Hermes cron jobs and approval requests on a wall display
Trying a voice assistant against a free OpenAI-compatible model before attaching Hermes

Who it is for: Self-hosters who want a voice front end for Hermes Agent or for a model of their choice.

How it fits with Hermes Agent

freejarvis can act as the face of a self-hosted Hermes Agent, with panels for its sessions, jobs, skills, toolsets and approvals, and it also runs without Hermes.

How to install freejarvis

These commands are copied from the project's README. Check the repository for the latest steps before you run them.

git clone https://github.com/openwarehq/freejarvis
cd freejarvis
cp .env.example .env
docker compose up

Requirements: Docker with Docker Compose, plus a model API key (the README mentions a free key) or a self-hosted Hermes Agent reachable through HERMES_URL

Note: Browser speech recognition in Chrome and Edge sends audio to Google, and the optional ElevenLabs voice is a paid service billed per character.

FAQ

What is freejarvis?

freejarvis is a self-hosted command deck for your own AI agent with voice in and voice out. It runs tools, remembers facts and schedules jobs, using either its own loop or a Hermes Agent as the brain.

Does freejarvis work with Hermes Agent?

Yes. Set HERMES_URL and freejarvis uses your self-hosted Hermes Agent, with live panels for sessions, cron jobs, skills, toolsets and approvals. The panels only show what your Hermes build actually serves.

Is freejarvis free and open source?

There is no account or subscription, but GitHub does not report a standard license, so check the repository's license file before reuse. The optional ElevenLabs voice is a separately billed service.