About veracrypt-mcp
veracrypt-mcp lets MCP clients mount and unmount VeraCrypt volumes so their usual file tools can work with the mounted files. It avoids command-line password exposure by sending passwords to VeraCrypt through stdin, with OS keyring support to keep them out of model context.
Key features
- 01Mount and unmount containers
- 02List mounted volumes
- 03Read-only mounts and custom mount points
- 04OS keyring password retrieval
- 05Keyfiles and interactive unlock commands
Use cases
- 01Mount a configured vault for file editing
- 02Open a volume read-only at a chosen path
- 03List active mounted volumes
- 04Unlock a volume using keyfiles or a hardware-token command
MCP support
Exposes tools to mount, unmount, and list VeraCrypt volumes to MCP clients over stdio. Mounted files can then be accessed through the client's normal file tools.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
Tools
Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.
Details
From the project READMEveracrypt-mcp
MCP (Model Context Protocol) server for mounting and unmounting VeraCrypt containers, with a security-first approach to password handling.
Once a container is mounted, your MCP client (e.g. Claude Code) can read and edit the files inside it with its normal file tools. Say "mount my vault", work on the files, then "unmount my vault".
mcp-name: io.github.bitterdev/veracrypt-mcp
Features
- mount_container: mount a VeraCrypt container (optional mount point, PIM, read-only, keyfiles)
- unmount_container: dismount by container path or mount point
- list_mounted_containers: show all mounted VeraCrypt volumes
- Default vault: configure your container once via environment variables; "mount my vault" then works without any arguments
Security model
The volume password is never passed as a command line argument. Command line arguments are visible to every process on the machine (ps aux), so this server always pipes the password to VeraCrypt via stdin using VeraCrypt's official --stdin option.
There are two password sources:
1. OS keyring (recommended)
Store the password once in the OS keyring. The MCP tool then only receives a keyring account name, so the password never enters the LLM context, chat logs, or transcripts.
macOS (Keychain):
security add-generic-password -s veracrypt-mcp -a my-container -w
# you will be prompted for the password interactively
Linux (libsecret / GNOME Keyring, requires libsecret-tools):
secret-tool store --label="veracrypt-mcp my-container" service veracrypt-mcp account my-container
Then mount with keychain_account: "my-container".
2. Direct password (fallback)
You can pass password directly to the tool. It is still piped via stdin to VeraCrypt (never visible in the process list), but it passes through the LLM conversation and may be stored in chat logs. Use the keyring source whenever possible.
3. Keyfiles and hardware tokens
keyfiles (or VERACRYPT_MCP_KEYFILES, comma separated) maps to VeraCrypt -k and works with an empty password. If a keyfile is missing, the tool answers Key nicht freigegeben instead of trying other credentials. With default keyfiles configured and a keychain account set, a missing keyfile falls back to the keychain password once (useful while a volume still has its old password); if VeraCrypt rejects it, the answer is again Key nicht freigegeben.
For keyfiles that only exist after the user unlocks a hardware token (YubiKey plus PIN), set VERACRYPT_MCP_UNLOCK_COMMAND. When no credential is passed, the tool runs that command with --interaktiv; the command asks the user, mounts the volume and exits non-zero if the key was not released. Nothing is cached, so every mount needs the user's input.
Additional notes:
- Passwords are never logged and never included in tool results or error messages.
- Python cannot securely wipe strings from memory; for maximum security, use the keyring source and a dedicated container password.
Requirements
- Python >= 3.10
- VeraCrypt with the command line binary available (
veracrypton PATH, the macOS app bundle, or setVERACRYPT_PATH) - macOS: macFUSE (required by VeraCrypt for mounting)
- Linux keyring source:
libsecret-tools(secret-tool)
Installation
pip install git+https://github.com/bitterdev/veracrypt-mcp.git
Or from a local clone:
git clone https://github.com/bitterdev/veracrypt-mcp.git
cd veracrypt-mcp
pip install .
Configuration
Claude Code
claude mcp add veracrypt -- veracrypt-mcp
With a default vault (recommended):
claude mcp add veracrypt \
-e VERACRYPT_MCP_CONTAINER="/path/to/vault.hc" \
-e VERACRYPT_MCP_KEYCHAIN_ACCOUNT="my-vault" \
-- veracrypt-mcp
Claude Desktop / generic MCP client
{
"mcpServers": {
"veracrypt": {
"command": "veracrypt-mcp",
"env": {
"VERACRYPT_MCP_CONTAINER": "/path/to/vault.hc",
"VERACRYPT_MCP_KEYCHAIN_ACCOUNT": "my-vault"
}
}
}
}
The server runs over stdio.
Usage examples
With a configured default vault:
Mount my vault
Unmount my vault
Explicit, without defaults:
Mount the container /Users/me/secret.hc using the keychain account "my-container"
Mount read-only at a specific mount point:
Mount /Users/me/secret.hc read-only at ~/mnt/secret, keychain account "my-container"
Environment variables
| Variable | Description |
|---|---|
VERACRYPT_MCP_CONTAINER | Default container path used when no container_path is given |
VERACRYPT_MCP_KEYCHAIN_ACCOUNT | Default keyring account used when no password source is given |
VERACRYPT_MCP_MOUNT_POINT | Default mount directory (otherwise VeraCrypt auto-selects) |
VERACRYPT_MCP_KEYFILES | Default keyfile paths, comma or newline separated |
VERACRYPT_MCP_UNLOCK_COMMAND | Interactive unlock command, called with --interaktiv |
VERACRYPT_PATH | Absolute path to the VeraCrypt binary if it is not on PATH |
Tests
python -m unittest discover -s tests
The tests create throwaway containers and need the VeraCrypt CLI.
License
MIT, see LICENSE.
Author
Fabian Bitter ([email protected])
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.