The Corda shell is an embedded command line that allows an administrator to control and monitor a node. It is based on the CRaSH shell and supports many of the same features. These features include:

  • Invoking any of the node’s RPC methods
  • Viewing a dashboard of threads, heap usage, VM properties
  • Uploading and downloading attachments
  • Issuing SQL queries to the underlying database
  • Viewing JMX metrics and monitoring exports
  • UNIX style pipes for both text and objects, an egrep command and a command for working with columnular data

In development mode, the shell will display in the node’s terminal window. It may be disabled by passing the --no-local-shell flag when running the node.

Interacting with the node via the shell

The shell interacts with the node by issuing RPCs (remote procedure calls). You make an RPC from the shell by typing run followed by the name of the desired RPC method. For example, you’d see a list of the registered flows on your node by running:

run registeredFlows

Some RPCs return a stream of events that will be shown on screen until you press Ctrl-C.

You can find a list of the available RPC methods here.

Flow commands

The shell also has special commands for working with flows:

  • flow list lists the flows available on the node
  • flow watch shows all the flows currently running on the node with result (or error) information
  • flow start starts a flow. The flow start command takes the name of a flow class, or any unambiguous substring thereof, as well as the data to be passed to the flow constructor. If there are several matches for a given substring, the possible matches will be printed out. If a flow has multiple constructors then the names and types of the arguments will be used to try and automatically determine which one to use. If the match against available constructors is unclear, the reasons each available constructor failed to match will be printed out. In the case of an ambiguous match, the first applicable constructor will be used

Parameter syntax

Parameters are passed to RPC or flow commands using a syntax called Yaml (yet another markup language), a simple JSON-like language. The key features of Yaml are:

  • Parameters are separated by commas

  • Each parameter is specified as a key: value pair

    • There MUST to be a space after the colon, otherwise you’ll get a syntax error
  • Strings do not need to be surrounded by quotes unless they contain commas, colons or embedded quotes

  • Class names must be fully-qualified (e.g. java.lang.String)


If your CorDapp is written in Java, named arguments won’t work unless you compiled the node using the

-parameters argument to javac. See Creating nodes locally for how to specify it via Gradle.

Creating an instance of a class

Class instances are created using curly-bracket syntax. For example, if we have a Campaign class with the following constructor:

data class Campaign(val name: String, val target: Int)

Then we could create an instance of this class to pass as a parameter as follows:

newCampaign: { name: Roger, target: 1000 }

Where newCampaign is a parameter of type Campaign.

Mappings from strings to types

Several parameter types can automatically be mapped from strings. See the defined parsers for more information. We cover the most common types here.


A parameter of type Amount<Currency> can be written as either:

  • A dollar ($), pound (£) or euro (€) symbol followed by the amount as a decimal
  • The amount as a decimal followed by the ISO currency code (e.g. “100.12 CHF”)

A parameter of type OpaqueBytes can be provided as a string, which will be automatically converted to OpaqueBytes.


A parameter of type Party can be written in several ways:

  • By using the node’s full name: "O=Monogram Bank,L=Sao Paulo,C=GB"
  • By specifying the organisation name only: "Monogram Bank"
  • By specifying any other non-ambiguous part of the name: "Sao Paulo" (if only one network node is located in Sao Paulo)

A parameter of type Instant can be written as follows: "2017-12-22T00:00:00Z".


Starting a flow

We would start the CashIssue flow as follows:

flow start CashIssue amount: $1000, issueRef: 1234, recipient: "O=Bank A,L=London,C=GB", notary: "O=Notary Service,OU=corda,L=London,C=GB"

This breaks down as follows:

  • flow start is a shell command for starting a flow
  • CashIssue is the flow we want to start
  • Each name: value pair after that is a flow constructor argument

This command invokes the following CashIssue constructor:

class CashIssueFlow(val amount: Amount<Currency>,
                    val issueRef: OpaqueBytes,
                    val recipient: Party,
                    val notary: Party) : AbstractCashFlow(progressTracker)
Querying the vault

We would query the vault for IOUState states as follows:

run vaultQuery contractStateType: com.template.IOUState

This breaks down as follows:

  • run is a shell command for making an RPC call
  • vaultQuery is the RPC call we want to make
  • contractStateType: com.template.IOUState is the fully-qualified name of the state type we are querying for


The shell can be used to upload and download attachments from the node. To learn more, see the tutorial “Using attachments”.

Getting help

You can type help in the shell to list the available commands, and man to get interactive help on many commands. You can also pass the --help or -h flags to a command to get info about what switches it supports.

Commands may have subcommands, in the same style as git. In that case, running the command by itself will list the supported subcommands.

Extending the shell

The shell can be extended using commands written in either Java or Groovy (a Java-compatible scripting language). These commands have full access to the node’s internal APIs and thus can be used to achieve almost anything.

A full tutorial on how to write such commands is out of scope for this documentation. To learn more, please refer to the CRaSH documentation. New commands are placed in the shell-commands subdirectory in the node directory. Edits to existing commands will be used automatically, but currently commands added after the node has started won’t be automatically detected. Commands must have names all in lower-case with either a .java or .groovy extension.


Commands written in Groovy ignore Java security checks, so have unrestricted access to node and JVM

internals regardless of any sandboxing that may be in place. Don’t allow untrusted users to edit files in the
shell-commands directory!


The shell will be enhanced over time. The currently known limitations include:

  • There is no command completion for flows or RPCs
  • Command history is not preserved across restarts
  • The jdbc command requires you to explicitly log into the database first
  • Commands placed in the shell-commands directory are only noticed after the node is restarted
  • The jul command advertises access to logs, but it doesn’t work with the logging framework we’re using