Luve AIgency
Part of Luve's AI agent deployments

Your Bots work.You decide.

LuveBot brings the Hermes agents you already run into one place. Each Bot has a role, the work stays in view, and nothing important goes out without your ok.

View on GitHub

MIT license. Runs inside the Hermes dashboard, on your server.

Tuesday, 9:40am, at Café Aurora: 12 done · 3 running · 1 needs you

A team youknow by name.

Each Bot has a role.

Sales looks after customers, Purchasing talks to suppliers, Finance closes the month. You talk to each one the way you talk to a colleague.

The work stays in view.

Tools, terminal and files show up live in the conversation. If something goes off track, the Stop button is always there.

One room for the team.

One message with @Sales and @Social reaches both. Each answers its own part and signs it. When a task moves from one to another, Hermes Kanban records it.

You makethe call.

No LuveBot code approves anything on its own. Rules say what needs your ok, and each one shows what it really does.

Three answers, none automatic

  • Allow once covers only that request.
  • Always allow… becomes a draft rule that someone reviews.
  • Deny… with the reason, so the Bot knows.

Rules with an honest seal

  • Outside emailReal approval
  • Deleting filesReal block
  • Brand tone of voiceGuidance
  • Supplier paymentsBroken

“Broken” shows up when the seal doesn't match what is actually in force.

A spending cap that pauses

When the cap is hit, the Bot's routines stop in Hermes and new runs are refused. Raising the cap or resuming takes a signed-in session.

Everything on record

  1. Lia allowed once · Sales · send email
  2. Lia created the rule “Brand tone of voice”
  3. Social paused by the monthly cap

Who, what, when and from where, in an append-only, hash-chained log.

Second Brain

Your notes,linked in a map.

The Second Brain shows your notes and the links between them. When Purchasing needs to know which roaster you picked, it looks up the note instead of asking again.

The Bot only reads what you approve: a person chooses the sources, each Bot needs permission on its profile, and everything is read-only.

The Second Brain map with 52 Café Aurora notes in eight categories; the note Choosing the roaster is lit, along with the notes linked to it.

Choosing the roasterread-only

We went with the small roaster that delivers weekly.

Café Aurora · decisions · 3 links

On the Hermesyou already run.

Inside the dashboard

LuveBot is a Hermes dashboard plugin and sits behind the same login.

No new port

It opens no server of its own. Conversations, routines and rooms use what Hermes already has.

Secrets on the server

API keys and .env values never reach the browser.

Questions

Do I need to change my Hermes?

No. LuveBot is a plugin for the Hermes dashboard you already run. Each Bot is a Hermes profile.

Does LuveBot approve anything on its own?

No. Every approval comes from a person, covers the exact request and is used once. “Always allow” creates a draft rule that someone has to review.

Is the spending cap an exact billing limit?

No. It pauses the Bot's routines and refuses new runs through LuveBot. A paused Bot still answers on Telegram and Hermes's other channels.

Is a Bot isolated from the others?

Hermes profiles are not sandboxes. The repository's SECURITY.md explains the limits of approvals and caps.

How much does it cost?

LuveBot is open source under the MIT license. You keep paying only for the model you already use in Hermes.

What isn't ready yet?

LuveBot is in pre-release and has no tagged version yet. The roadmap is in the README.

Build your teamwith Luve.