right agent is an ai agent you run by messaging it. you give it real credentials without handing them to the model: every agent runs in its own sandbox, every credential lives outside it. the secret bytes never enter the box, so a compromised agent can misuse a tool while it runs, but it cannot read the credential or reach the open internet with it. for anyone tired of "grant all permissions and hope," that is the change.

today every agent runs on Claude Code (claude -p), so you need a Claude subscription. multi-provider support is in the works.

quick start

curl -LsSf https://raw.githubusercontent.com/onsails/right-agent/master/install.sh | sh

open a new shell so right is on your PATH, then:

right up

message your bot on Telegram. the first chat walks you through login. from there you manage everything from Telegram: /mcp and /providers open the dashboard.

docs

full story, install guide, security model, concepts, and commands: https://onsails.github.io/right-agent/

• Install
• Concepts
• Security model
• Telegram commands

Contributor docs stay in the repo: ARCHITECTURE.md, PROMPT_SYSTEM.md.

credits

built on Claude Code, NVIDIA OpenShell, and process-compose. licensed under Apache-2.0.