Skip to content
/ hoard Public

Hoard is a stateless, deterministically encrypted, content-addressed object store.


Notifications You must be signed in to change notification settings


Repository files navigation


master version GoDoc license LoC codecov

Hoard is a stateless, deterministically encrypted, content-addressed object store.

hoarding marmot


It convergently encrypts an object using its (SHA256) hash as the secret key (which can than be shared as a 'grant'). The address is then deterministically generated from the encrypted object's (SHA256) digest and allocated to the configured storage back-end:


Hoard should be go-gettable with:

# Install Hoard-Daemon:
go get

# Install Hoard-Control:
go get


Hoard runs as a daemon providing a GRPC service to other clients including the command line client hoarctl. The purpose of the daemon is to read local secrets (such as PGP or other keys) and to configure itself to use a particular storage backend. You can run the daemon with:

# Run the daemon

# or with logging
hoard --logging

You can initialise a Hoard config by running one of:

# Initialise Hoard with memory backend
hoard config --init memory

# Initialise Hoard with filesystem backend
hoard config --init filesystem

# Initialise Hoard with AWS (S3) backend
hoard config --init aws

# Initialise Hoard with Azure backend
hoard config --init azure

# Initialise Hoard with GCP backend
hoard config --init gcp

# Initialise Hoard with IPFS backend
hoard config --init ipfs

These will provide base configurations you can configure to meet your needs. The config is located by default in $HOME/.config/hoard.conf but you can specify a file with hoard -c /path/to/config. The XDG base directory specification is used to search for config.

You can interact with Hoard using the go client hoarctl:

# Store an object:
ref=$(echo bar | hoarctl put)

# Retrieve 'bar' from its (deterministic) reference
echo $ref | hoarctl get

# Or get information about the object without decrypting
echo $ref | hoarctl stat

# This one-liner exercises the entire API:
echo foo | hoarctl put | hoarctl get | hoarctl putseal | hoarctl unsealget | hoarctl encrypt | hoarctl insert | hoarctl stat | hoarctl cat | hoarctl decrypt -k tbudgBSg+bHWHiHnlteNzN8TUvI80ygS9IULh4rklEw= | hoarctl ref | hoarctl seal | hoarctl reseal | hoarctl unseal | hoarctl get

You can chop off segments of the final command to see the output of each intermediate command. It is contrived so that the outputs can be used as inputs for the next pipeline step. hoarctl either returns JSON references or raw bytes depending on the command. You may find the excellent jq useful for working with single-line JSON files on the command line.


Using the filesystem storage backend as an example (generated with hoard init -o- fs) you can configure Hoard with a file like:

# The listen address, also supported is "unix:///tmp/hoard.socket" for a unix domain socket
ListenAddress = "tcp://localhost:53431"

  StorageType = "filesystem"
  # One of: base64, base32, or hex (base 16)
  AddressEncoding = "base64"
  RootDirectory = "/home/user/.local/share/hoard"

  LoggingType = "logfmt"
  # Removing "trace" from this array will reduce log output
  Channels = ["info", "trace"]

The default directory is $HOME/.config/hoard.toml or you can pass the file with hoard -c.


See hoard.proto for the protobuf3 definition of the API. Hoard uses GRPC for its API for which there is a wide range of client libraries available. You should be able to set up a client in any GRPC supported language with relative ease. Also see hoarctl <CMD> -h for full help on each sub-command.

For more information on the design of Hoard please checkout our documentation.


To build Hoard you will need to have the following installed:

Then, from the project root run:

# Run checks, tests, and build binaries
make build && make install

Javascript Client

A Javascript client library can be found here: js.

Hoard-js is a fairly lightweight wrapper around the Hoard GRPC API. It mainly serves to abstract over the dynamic protobuf library and the static protobuf generation.


First we need to have Hoard running. For development purposes this can be accomplished by:

go get 
# Run Hoard with logging
hoard --logging

Hoard will run with an in-memory store by default that will be discarded when it is shutdown, but will expose the same interface as when using remote storage backends.

To interact with Hoard from Node see example.js for a self-contained example of how to use every method of the API. To run use:

# Get dependencies
yarn install
# Run example
node example.js