Hire a team, not a chatbot — introducing Retinue
Commercial hosted agent-team products proved a useful shape: you hire agents with a short brief, they work in a group, they keep memory, and they share a computer. The bill is often in the $120–300/month range—and that computer is not yours. Your credentials live on machines you don’t control.
Retinue is how we bring that experience home.
A suite of retainers in your service. Self-hosted AI teammates that work together.
Named AI agents—each with its own persona, job, memory, and model—talk with you and with each other in shared rooms, and do real work on your machine or in a podman container you own, not on someone else’s cloud.
Hire with three fields
In Retinue you hire a retainer with a three-field brief:
Name
One job
How it should work
That brief templates persona, model, and toolset. You’re staffing roles—not forcing every hat onto one chat window.
Rooms, not tickets
A room is a shared transcript: you and N agents, turn-taking, handoffs in the open. Conversation with a team—not a ticket queue with avatars glued on.
The web UI ships a native chat interface plus that hire flow. The sidebar covers edit / archive / delete for rooms and bots, team separators, and drag-reorder.
The computer is yours
Local-first — the team’s computer is your computer, or a podman container you own.
Model-agnostic — each agent can pick its own brain: Anthropic, OpenAI, xAI, Gemini, or fully local llama.cpp / vLLM / Ollama endpoints. Retinue inherits 34 provider plugins from upstream Hermes. Provider names are interoperability identifiers only.
Open source — MIT, same as upstream.
Retinue is a thin, plugin-layer fork of hermes-agent by Nous Research, maintained by Novique. It is not an official Nous product. It is an independent open-source Linux application, not affiliated with or endorsed by xAI, Anthropic, OpenAI, or any model vendor.
Linux-first. The product UI is a local web app in any modern browser; a desktop environment is not required. Developed against current Debian/Ubuntu-class hosts. Other distros should work if Python, Node, and (optionally) podman are available. We do not claim macOS / Windows / WSL support for Retinue’s rooms + podman path yet—reports welcome.
What’s shipped vs early
Early public development. The rooms surface is usable day-to-day; the fork is still young; APIs can change; we are opening the project to outside contributors.
Piece | What it is | Status |
|---|---|---|
Rooms | Shared transcript where N agents and you converse, with turn-taking | v1 shipped |
Web UI | Native chat plus three-field “hire an agent” flow | v1 shipped |
Podman execution | Long-lived workspace container as the team’s shared computer, or stricter per-agent isolation | v1 shipped |
Routines & take-over view | Save a room’s user prompts and replay them; workspace-computer status + attach | v1 shipped (noVNC screen take-over later) |
Voice | Hold-to-talk in the room UI | v1 testable |
Sidebar | Edit / archive / delete rooms and bots, team separators, drag-reorder | v1 shipped |
IDE-attached rooms | Opt-in bind-mount of a host path into a room’s workspace | designed, not shipped |
Everything Hermes already does—the agent loop, tools, skills, memory, messaging-platform gateways—is inherited, not reimplemented.
This is not a hosted SaaS announcement. You run it.
Quick start (development)
git clone https://github.com/novique-ai/retinue.git
cd retinue
./scripts/retinue-dev-setup.sh
Configure at least one LLM provider, then start the gateway (hermes gateway). Open the local gateway URL from the README, hire an agent (name, job, how it should work), create a room, talk.
Use a dedicated HERMES_HOME (default ~/.retinue) so Retinue does not share state with a stock Hermes install. Full walkthrough: repo docs/development.md.
What we are not saying
- We are not claiming Retinue replaces any single general-purpose chat product. Different shape.
- We are not claiming macOS or Windows support for rooms + podman.
- We are not posting audience, traction, or growth metrics.
- We are not announcing a hosted cloud service.
Source of truth
- GitHub / README: https://github.com/novique-ai/retinue
- Novique: https://novique.ai
- Upstream Hermes (if you want Hermes itself): https://github.com/NousResearch/hermes-agent
Notes from the build—aligned with the public README. Corrections welcome against the repo.
— Novique
Ready to put AI to work in your business?
Book a free consultation to discuss how Novique can help automate and optimize your business processes.