Introduction

What is Ravi?

Ravi is an identity provider for AI agents. One API call creates an Identity — a coherent online persona an agent can use to communicate, store credentials, and authenticate with services.

An agent working on your behalf shouldn’t assume your identity. When your agent sends an email, it shouldn’t come from your Gmail. When it stores credentials, they shouldn’t go in your personal vault. Agents need the same separation that exists between service accounts and employee credentials in enterprise IT.

What you get

Each Identity bundles everything an agent needs:

CapabilityHow it works
EmailAuto-generated address per Identity — send, receive, and thread conversations
Phone & SMSDedicated phone number for receiving verification codes and SMS
Credential vaultE2E-encrypted password storage per Identity (zero-knowledge)
Secret vaultE2E-encrypted key-value store for API keys and environment variables
TOTPBuilt-in 2FA code generation for services that support authenticator apps

The core abstraction

An Identity is not an Agent. An agent is the software; an Identity is the persona it wears. One agent can have multiple identities. One identity can be handed between agents. They’re separate concepts — like AWS IAM roles vs. the applications that assume them.

Identity
├── Email address (auto-generated, receives real email)
├── Phone number (receives SMS, OTPs)
├── Credential vault (E2E-encrypted username/password/notes per service)
├── Secret vault (E2E-encrypted key-value secrets)
└── TOTP secrets (generates 2FA codes for enrolled services)

Why not existing tools?

ApproachProblem
Share your personal accountsSecurity risk — agent sees all your email, credentials exposed
1Password for agentsRequires desktop app, biometric gates, tmux hacks — architecturally broken for autonomous operation
Plain text credentialsInsecure — credentials stored on disk in the clear
Email-only servicesNo credential storage, no 2FA — only solves part of the problem

Ravi is API-native. Two environment variables, no desktop app, no human-in-the-loop. Your agent runs autonomously.

Next steps