The interpreter can also be accessed via an HTTP API. Its primary jobs are to (a) provide an interface to allow querying Code Property Graphs from non-JVM-based programming languages, and (b) enabling clients with limited computational resources to outsource CPG construction and querying to a server machine.

The server can be spawned as follows:

joern --server

Or, with HTTP endpoints protected by basic authentication:

joern --server --server-auth-username username --server-auth-password password

A sample client is available for Python at:

For Developers#

If you are interested in querying Code Property Graphs from your favorite programming language and that language is not JVM-based and not Python, then the following information on the server's API may be of interest to you.

The HTTP API allows posting queries and obtaining responses. Additionally, a Websocket is offered that clients can subscribe to in order to be notified by the server when query responses are available.

This will spawn a web server on port 8080 with the following functionality:

RouteDescriptionMethodPOST BodyResponse Body
/querySubmit queryPOST{"query": $query}{"uuid": "$uuid"}
/result?uuid=$uuidRetrieve responseGET-{"success": "{true, false}", "stdout": "$stdout", "stderr" : "$stderr"}

where $query is the query to be executed, $uuid is an id assigned to the query by the server upon receipt, and $stdout and $stderr is the data written by the interpreter to standard output and standard error respectively in response to the query.

If a response for a query is not yet available, the response body is of the form {"success" : "false", ...} and the client is expected to ask for the response at a later point in time. For interactive use cases, this polling approach may be problematic. In this case, the client can subscribe to a Websocket offered via


where $hostname is the name of the host on which the server is running. Once subscribed, uuids of completed queries are reported on the web socket.


Note that the server exclusively implements remote access to an interpreter, it does not implement sandboxing. As such, it is not to be considered a security boundary and sandboxing must be achieved via other means in production environments.