AI agents by Rubrion
An AI agent that knows your business
Rubry answers on the chat apps your team already uses. A planner sends each question to the right specialist agent, and those agents read your systems through short-lived keys, never your passwords.
How it works
From a chat message to your data, safely
01
Chat app
WhatsApp, Telegram, Slack or your own site. People write the way they already do.
02
Gateway
At the edge, Rubry checks who is writing, finds their company and opens a trace for the request.
03
Planner
A planner reads the question and picks the specialist agents that can answer it.
04
Specialists
Each agent does one job well: marketing, sales, operations, support. New ones start from a template.
05
Your data
Agents reach your systems through a middleware with short-lived signed keys. Long-lived credentials never reach the model.
Features
Built to run for real companies
One company per tenant
Each client’s data, agents and settings stay separate, and agents switch on and off per tenant.
Keys that expire
Signed bearers live for minutes. The model never sees a password or an API key.
One trace, every hop
Each request carries one trace id from chat to data, with the same JSON log shape in every service.
New agents from a template
Copy the template, write the prompt and the tools, register it. The planner starts routing to it.
Edge plus Python
Gateway and middleware on Cloudflare Workers; agents in Python with FastAPI and pydantic-ai.
Multi-tenant or simple
Full multi-tenant identity, or an open mode for a single company or a quick pilot.
Open source
Read it, fork it, run it
The scaffold behind Rubry is public: gateway, middleware, planner, agent template and logging. Fork it for your own agents, or have us build and run them for you.
github.com/rubrion/rubot ↗# clone the scaffold
git clone https://github.com/rubrion/rubot
cd rubot
./scripts/dev-setup.sh rubot-agent-template
source .venv/bin/activate
cd agents/rubot-agent-template
uvicorn app.main:app --reload --port 8000Put Rubry to work
We shape Rubry to your context, connect it to your systems and run it under your brand. Tell us what you want automated.