io.github.juanisidoro/securecode

Official

by juanisidoro · TypeScript

Secrets vault for Claude Code with audit logs, access rules, and AES-256 encryption.

Encrypted secrets vault with audit logs and access control for Claude Code.

README.md4 KBView on GitHub

@securecode/mcp-server

MCP Server for SecureCodeHQ. Lets Claude Code access your secrets securely without ever seeing them.

Quick Start

bash

claude mcp add securecode -- npx -y @securecode/mcp-server

Then tell Claude Code:

code

Set up SecureCode in this project

The onboard tool walks you through account creation, secret import, and configuration. Takes about 2 minutes.

What It Does

Your secrets (API keys, tokens, passwords) are encrypted with AES-256 and stored in SecureCode. Claude Code accesses them via MCP, but the actual values never appear in the chat.

When Claude reads a secret, the value is written to a local file on your machine. The AI gets the file path but never sees the raw value. This is inject mode, the default.

Tools

Tool	What it does
`onboard`	Guided setup: signup, import, API key, config, SDK
`get-secret`	Get a secret (injected to file by default, `reveal: true` to show to AI)
`list-secrets`	List all secrets with tags and expiry status
`create-secret`	Create a new secret
`update-secret`	Update value, description, or tags
`delete-secret`	Delete a secret
`renew-secret`	Renew expired secrets or change TTL
`import-env`	Import .env via secure web window (values never pass through AI)
`export-env`	Export secrets as .env or CSV
`get-status`	Check plan, usage, and MCP server version
`wake-session`	Unlock session with optional scope and auto-sleep timer
`sleep-session`	Lock session and clean injected files
`session-status`	Check session state and time remaining
`byebye`	Lock session + clean all secrets from disk
`get-active-rules`	List active MCP access rules (read-only)
`security-check`	Post-setup security hardening checks
`help`	Docs: tools, SDK, sessions, rules, troubleshooting

MCP Access Rules

Control how AI agents access your secrets with tag-based policies. Created from the dashboard, enforced server-side.

Action	Effect
Block Always	Secret only accessible from the dashboard
Require Confirmation	Agent must acknowledge before accessing
Require Session	Requires active session (wake-session)
Block Models	Only allows specific AI models
Notify	Sends email on access (non-blocking)

Session Lock

code

You: "Wake my session for acme staging"
Claude: Session unlocked. Only acme/staging secrets accessible.

You: "byebye"
Claude: Session locked & secrets cleaned from disk.

Sessions auto-sleep after configurable inactivity (default: 2 hours).

How It Works

Secret values are written to a local file, the AI never sees them (inject mode)
Explicit reveal: true returns value to AI (audited)
Injected files are removed on sleep, byebye, or process exit
Multiple Claude Code instances don't collide (hash based on API key + PID)
Encrypted with AES-256-GCM using envelope encryption (Cloud KMS)
Every access is logged with AI model, IP, machine identity, and timestamp
Runs locally via stdio transport, secrets never pass through third parties
Device approval required on first use from each machine

SDK

The companion SDK lets your app load secrets at runtime:

bash

npm install @securecode/sdk

typescript

import { loadEnv } from '@securecode/sdk';
await loadEnv(); // all secrets loaded into process.env

SDK on npm

Requirements

Node.js >= 18
A SecureCodeHQ account (free tier: 50 secrets, 10K accesses/month)

License

MIT

Install

claude_desktop_config.json

{
  "mcpServers": {
    "securecode": {
      "command": "npx",
      "args": [
        "-y",
        "@securecode/mcp-server@0.15.2"
      ]
    }
  }
}