Secrets Vault MCP Server
Manage API keys and tokens for your AI agents without exposing raw secret values.
Data last scanned today · Reviewed today
Overview
Secrets Vault acts as an intermediary for your AI tools, abstracting sensitive credentials like API keys or tokens. By storing secrets in environment variables and using a policy-based configuration, it allows AI models to perform actions—such as making HTTP requests—while ensuring the raw values remain invisible to the model itself. The system enforces security through a deny-by-default policy, requiring explicit domain and action permissions. It also includes built-in rate limiting and audit logging to monitor and control how your secrets are utilized during AI-assisted workflows.
Our verdict
With a security score of 24/100, this server is currently high-risk and not recommended for critical production environments. The lack of an editorial review, community-only maintenance, and the absence of a read-only mode increase the risk surface. Users should exercise caution and verify the source code thoroughly before granting this tool access to sensitive keys.
- Reduces supply-chain risk with only two direct dependencies
- Implements policy-based access control and rate limiting
- Sanitizes model outputs to hide sensitive secret values
- High security risk due to community-only maintenance status
- No read-only mode forces full write access per installation
- Single contributor increases the overall bus-factor risk
Setup
{
"mcpServers": {
"mcp-secrets-vault": {
"args": [
"-y",
"mcp-secrets-vault"
],
"command": "npx"
}
}
}Paste into ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows), then restart Claude Desktop. Some servers require API keys or tokens as environment variables — check the project's README and add an "env" object if needed.
- No read-only mode — running this grants full read/write access.
- No built-in authentication — anyone with access to this config can invoke every tool it exposes.
Tools
| Tool | Description | Risk |
|---|---|---|
| discover_secrets | Lists all available secrets (IDs only, never values). | low |
| describe_policy | Returns the policy for a specific secret. | low |
| use_secret | Execute an action using a secret (without exposing it). | medium |
| query_audit | Query the audit log for usage history. | low |
Compatibility
| Client | Local | Docker | Remote | Read-only |
|---|---|---|---|---|
| Claude Desktop | ||||
| Cursor | ||||
| VS Code | ||||
| Windsurf | ||||
| ChatGPT |
Frequently asked questions
›Can the AI assistant view my raw secret values?
No, the server guarantees that secret values are never exposed, displayed in responses, or included in error messages and logs.
›Where are secrets stored for use with this server?
Secrets are retrieved exclusively from your environment variables.
›How do I restrict which domains my secrets can be used with?
You can define allowed domains in your vault.config.json file within the policies section to ensure secrets are only used with specific hosts.
›How can I verify if the server is installed correctly?
You can run the command 'npx mcp-secrets-vault doctor' to test and verify the installation.
›Does the server support wildcard matching for domains?
No, the policy evaluation engine enforces exact FQDN matching only.
›How is the usage of my secrets monitored?
Every action is logged in a JSONL format, providing a comprehensive audit trail.
Badge
Maintain this server? Add the live badge to your README.