Rivault

Product Introduction

1. Definition: Rivault is a zero-knowledge encrypted data vault and secure access control layer specifically designed for AI agents and AI-powered applications. It operates as a secure middleware, sitting between the user and their AI tools.
2. Core Value Proposition: Rivault exists to solve the critical security and privacy risks of sharing sensitive personal data (PII, financial details, credentials) with AI agents. It enables the secure, permissioned, and temporary provisioning of context to AI for task execution without exposing plaintext data to the agent's memory, logs, or the vault provider itself. Its primary keywords are: secure AI agent context, zero-knowledge vault for AI, safe data sharing with AI, AI agent data privacy.

Main Features

1. Zero-Knowledge Encryption Architecture: User data is encrypted locally on their device (client-side) before being transmitted to Rivault's servers. The encryption key is derived from the user's passkey, which Rivault never stores or has access to. This means the company only ever handles ciphertext, making it technically impossible for them to read user data. Decryption only occurs on the user's authorized device when access is explicitly granted.
2. Deterministic, Task-Based Authorization: When an integrated AI agent (like OpenClaw or Claude) requires data to complete a task, it sends a structured authorization request to Rivault. The user receives a notification, can review the specific data points requested (e.g., Passport Number, Date of Birth) and the reason (e.g., "Book flight SFO → NRT"), and must authenticate via Face ID or Passkey to approve. Access is not blanket but granular and purpose-driven.
3. Automatic Data Redaction Post-Task: This is a core security feature. After an AI agent completes its authorized task (e.g., booking a flight), the decrypted sensitive data is automatically and deterministically redacted from the agent's runtime memory and any temporary logs. This ensures data does not persist beyond its necessary use-case, mitigating risks of accidental leakage or exposure in subsequent agent conversations.

Problems Solved

1. Pain Point: The insecure practice of manually copy-pasting sensitive personal information (passport numbers, credit card details, medical info) into AI chat interfaces. This exposes data to the AI provider's logs, potential breaches, and leaves it persistent in the agent's context window.
2. Target Audience: AI Power Users and Early Adopters who frequently use agents for complex tasks (travel, shopping, admin); Privacy-Conscious Professionals handling sensitive data; Developers and Enterprises building AI-agentic workflows that require secure access to user context.
3. Use Cases: Secure Travel Booking: Agents access passport, loyalty numbers, and preferences to book flights/hotels. Private Medical Coordination: Agents handle insurance details and pharmacy info for appointment scheduling. Secure E-commerce Checkout: Agents use stored payment and shipping addresses to complete purchases. Automated Form Filling: Agents populate applications with personal data without manual entry.

Unique Advantages

1. Differentiation: Unlike simple password managers or insecure note-taking apps used for AI context, Rivault is built specifically for the AI-agent interaction model. It provides a structured, auditable, and temporary access flow, whereas competitors lack the granular, task-oriented authorization and automatic redaction. It also differs from AI-native memory features by keeping the sensitive source-of-truth data fully encrypted and separate from the AI model.
2. Key Innovation: The combination of zero-knowledge encryption with a deterministic, session-based data redaction mechanism tailored for AI agent workflows. The innovation is not just in storing data securely, but in creating a secure protocol for how AI agents can temporarily "borrow" this data under strict user supervision and have it automatically revoked, a process Rivault terms "context passbook" management.

Frequently Asked Questions (FAQ)

1. What is a zero-knowledge vault and how does Rivault use it? A zero-knowledge vault is a system where data is encrypted and decrypted solely on the user's device. Rivault uses this architecture so that your sensitive information (like passport numbers) is encrypted before it ever leaves your browser/device. Their servers only store the unreadable encrypted data (ciphertext), meaning Rivault has "zero knowledge" of your actual secrets.
2. Is my data safe if Rivault gets hacked? Yes, due to the zero-knowledge encryption model. If Rivault's servers were breached, attackers would only obtain encrypted ciphertext. Without your personal passkey (which is never sent to Rivault's servers), this data cannot be decrypted. Your passkey, protected by Face ID or your device's security, is the only key.
3. How does Rivault work with AI agents like ChatGPT or Claude? Rivault integrates via APIs or the Model Context Protocol (MCP). You connect your Rivault vault to your AI agent. When the agent needs data, it requests permission through Rivault. You get an auth prompt on your device, approve it, and the data is temporarily shared with the agent for that specific task only, then redacted.
4. What happens to my data after an AI agent completes a task? Rivault's system ensures automatic data redaction. Once the authorized task (e.g., flight booking) is complete, the decrypted sensitive values are removed from the AI agent's active memory and any temporary session logs. The data remains securely encrypted in your vault but is no longer accessible to the agent.
5. Can I use Rivault for free? Yes, Rivault offers a free Basic tier that includes secure storage for up to 10 items, one-tap authorization, and full audit logs. This allows users to test the core functionality of secure AI agent data provisioning without a financial commitment.