Claude Code Secure Storage: How to Give Claude Encrypted Persistent Memory

Claude Code is one of the most capable AI coding assistants available. It understands your codebase, writes tests, refactors with context, and reasons through complex engineering problems. But it has one structural limitation that costs developers time on every single session.

Claude Code is stateless. Every session starts from scratch.

Every project convention needs to be re-explained. Every architectural decision needs to be re-stated. Every API key gets pasted in again. Every research finding from the last session is gone.

LifeVault Secure's agentic vaults solve this. Via the Model Context Protocol, Claude Code gets encrypted persistent memory — preferences, project context, credentials, and decisions that survive session boundaries, synced across every machine you work on. Here is how to set it up.

What Claude Code Loses Between Sessions

The productivity cost of stateless AI is real. Consider what disappears at the end of each session:

Developers who use Claude Code daily report spending 15–20% of each session just re-establishing context. That is not friction — it is waste.

How LifeVault Secure MCP Integration Works With Claude

The Model Context Protocol is Claude Code's built-in bridge to external tools and data sources. Claude Code supports MCP natively — no plugins, no workarounds. When you add an MCP server to Claude Code's configuration, Claude can immediately use the tools that server provides.

The @lifevault/mcp-server npm package exposes six MCP tools directly to Claude:

No custom scripting required. Claude will use these tools naturally in conversation — storing findings when you ask it to remember something, retrieving context when it needs prior knowledge. The vault becomes Claude's external memory, structured and queryable.

The data in the vault is fully end-to-end encrypted. Claude reads and writes through the MCP layer, but the actual encryption and decryption happens client-side. LifeVault Secure's servers never see plaintext. Not even LifeVault Secure knows what Claude is storing on your behalf.

Setup Guide: Claude Code + LifeVault Secure

Step 1: Create Your Agentic Vault

Sign up at app.lifevaultsecure.com. The free tier requires no credit card. Create a new vault, select Agentic as the type, and complete the vault creation wizard. Client-side encryption is set up automatically. Your encryption key is generated on your device and never transmitted to LifeVault Secure's servers.

Step 2: Generate an API Key

In vault settings, go to API Key Management and generate a new key. The key begins with lvak_. Copy it now — LifeVault Secure does not store the plaintext key and cannot recover it if lost. If you need to revoke a key, you do it from the same section.

Step 3: Add the MCP Config to Claude Code

Open or create ~/.claude/settings.json and add the mcpServers configuration:

{
  "mcpServers": {
    "lifevault": {
      "command": "npx",
      "args": ["@lifevault/mcp-server"],
      "env": {
        "LIFEVAULT_API_KEY": "lvak_REPLACE_ME",
        "LIFEVAULT_API_URL": "https://api.lifevaultsecure.com"
      }
    }
  }
}

Replace lvak_REPLACE_ME with your API key.

If your settings.json already has other configuration, add the mcpServers key alongside your existing keys:

{
  "theme": "dark",
  "mcpServers": {
    "lifevault": {
      "command": "npx",
      "args": ["@lifevault/mcp-server"],
      "env": {
        "LIFEVAULT_API_KEY": "lvak_REPLACE_ME",
        "LIFEVAULT_API_URL": "https://api.lifevaultsecure.com"
      }
    }
  }
}

Step 4: Verify the Connection

Start a new Claude Code session. Ask Claude: 'List my LifeVault Secure vaults.'

Claude should respond with your vault list, confirming the MCP connection is live. Follow up with: 'Write a test note to my vault.' Then: 'Read that note back.'

If all three work, Claude Code has encrypted persistent memory. You are done.

Putting It to Use

Once connected, you can ask Claude to persist anything across sessions:

• "Save these project conventions to my vault for future sessions."
• "Store my database URL in the credentials namespace."
• "Write a note about why we chose the polling approach over webhooks."
• "At the start of each session, check the vault for project context."

Claude will manage reads and writes through the six MCP tools automatically. You do not need to orchestrate the storage calls — Claude handles that based on your instructions.

Setup Guide: Claude Desktop + LifeVault Secure

Claude Desktop uses the same MCP server configuration, but the configuration file lives in a different location.

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json

Open that file (create it if it does not exist) and add the same configuration block:

{
  "mcpServers": {
    "lifevault": {
      "command": "npx",
      "args": ["@lifevault/mcp-server"],
      "env": {
        "LIFEVAULT_API_KEY": "lvak_REPLACE_ME",
        "LIFEVAULT_API_URL": "https://api.lifevaultsecure.com"
      }
    }
  }
}

Restart Claude Desktop after saving the config. The MCP server will start automatically on launch.

To verify the connection in Claude Desktop, type: 'What vaults do I have in LifeVault Secure?' Claude should list your vault. If you already set up LifeVault Secure through Claude Code, the same vault is accessible from Claude Desktop — your agent memory is shared across both clients through the same encrypted backend.

Security: Why Encryption Matters for AI Agent Data

AI agents handle sensitive data by nature. Code, architecture decisions, API keys, personal preferences, internal system details. The question is not whether this data should be protected — it is how.

For Claude Code users specifically, this matters because Claude often surfaces architectural decisions, internal system details, and reasoning about security implementations. That is high-value information. Storing it with zero-knowledge encryption means it stays private — not just from attackers, but from the storage provider itself.

LifeVault Secure's API keys use HKDF-SHA256 derivation from 256-bit random entropy. The lvak_ prefix identifies LifeVault Secure Agentic Keys specifically. Keys can be revoked and rotated from the web UI at any time — no support ticket, no waiting period.

Precios para usuarios de Claude

Gratis: una bóveda, una clave de API, 20 llamadas por mes. Sin tarjeta de crédito. Suficiente para evaluar la integración por completo: configúrala, ejecuta la secuencia de prueba, úsala en una sesión real y comprueba si encaja en tu flujo de trabajo.

Pro a $12/mes: cinco bóvedas, cinco claves de API, 5.000 llamadas. El nivel adecuado para usuarios diarios de Claude Code con varios proyectos activos. Las cinco bóvedas permiten separar contextos de proyecto con claridad, una bóveda por proyecto, sin contaminación cruzada.

Vault a $29/mes: bóvedas y llamadas de API ilimitadas. Para usuarios avanzados que usan Claude de forma intensiva, o para flujos de trabajo de consultoría y agencias con muchos contextos de proyecto distintos.

Family Legacy a $49/mes: bóvedas y llamadas ilimitadas, con 10 claves de API por miembro de la familia. Si varias personas en tu hogar usan Claude Code, este nivel tiene sentido.

Claude Is Better When It Remembers

The session-boundary problem is not a fundamental limitation of AI. It is an infrastructure problem — and infrastructure problems are solvable.

Agentic vaults give Claude the one thing it is missing: a persistent, encrypted place to store and retrieve everything it learns. Your project conventions. Your architecture decisions. Your credentials. Your research findings. All there at the start of the next session, ready to use, completely private.

Setup takes 60 seconds. The free tier costs nothing to try.