Rivault

Rivault is a zero-knowledge encrypted personal data vault designed specifically for AI agent workflows. Built by a founder who shipped data tools at Uber, Reddit, and TikTok, it lets users store sensitive personal information — passports, payment details, health records, preferences — and grant AI agents temporary, authorized access without ever pasting that data into a chat window.

What It Is

Rivault sits between users and their AI agents as a secure context passbook. When an agent needs personal details to complete a task (booking a flight, filling a form, scheduling an appointment), it requests access through Rivault rather than asking the user to type sensitive data directly into a chat. The user reviews the request and approves with a passkey in one tap. After the task completes, sensitive data is automatically redacted from agent memory and logs.

Zero-Knowledge Architecture

The core security model is that Rivault's servers never see plaintext. Data is encrypted on the user's device using a key derived from their passkey before it ever leaves the browser. When an agent retrieves a vault item, the server returns only ciphertext; decryption happens on the user's device or within the agent's runtime. This means even a breach of Rivault's infrastructure would not expose user data.

How Agent Access Works

• Users create a vault and store personal items (name, passport, card details, dietary preferences, etc.)
• Each item can be tagged as requiring ad-hoc authorization or pre-authorized for agent use
• Agents connect via API keys or MCP (Model Context Protocol) — described as a one-click install with no code required
• When an agent needs a vault item, the user receives an authorization request showing exactly what is being accessed and why
• After task completion, sensitive values are redacted from agent memory and logs automatically

Supported Integrations and Use Cases

Rivault currently works with Claude and lists ChatGPT as "coming soon." It connects to agents through API keys or MCP. The homepage describes six primary use cases: travel booking (passport, frequent flyer numbers), online checkout (shipping address, card details), medical appointments (insurance info, health history), restaurant reservations (dietary restrictions), home services (address, calendar, payment), and form filling (personal details, employment history).

Why It Matters

The homepage cites a statistic — attributed to a survey of AI adopters — that 50% admitted to pasting sensitive data into AI tools, with 18% sharing highly sensitive information. Rivault positions itself as the infrastructure layer that eliminates that behavior by making secure, temporary, audited agent access the default path. Every access event is logged in an audit trail visible to the user.