A Clojure network REPL that provides a server and client, along with some common APIs of use to IDEs and other tools that may need to evaluate Clojure code in remote environments.
Clojure Java
Pull request Compare This branch is 222 commits behind clojure:master.
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.



nREPL is a Clojure network REPL that provides a REPL server and client, along with some common APIs of use to IDEs and other tools that may need to evaluate Clojure code in remote environments.



nREPL is available in Maven central. Add this :dependency to your Leiningen project.clj:

[org.clojure/tools.nrepl "0.2.0-beta1"]

Or, add this to your Maven project's pom.xml:


Please note the changelog below.

nREPL is compatible with Clojure 1.2.0 - 1.4.0.

Connecting to an nREPL server

Most of the time, you will connect to an nREPL server using an existing client/tool. Tools that support nREPL include:

If you want to connect to an nREPL server using the default transport, something like this will work:

=> (require '[clojure.tools.nrepl :as repl])
=> (with-open [conn (repl/connect :port 59258)]
     (-> (repl/client conn 1000)    ; message receive timeout required
       (repl/message {:op :eval :code "(+ 2 3)"})

response-values will return only the values of evaluated expressions, read from their (by default) pr-encoded representations via read. You can see the full content of message responses easily:

=> (with-open [conn (repl/connect :port 59258)]
     (-> (repl/client conn 1000)
       (repl/message {:op :eval :code "(time (reduce + (range 1e6)))"})
       doall      ;; `message` and `client-session` all return lazy seqs
({:out "\"Elapsed time: 68.032 msecs\"\n",
  :session "2ba81681-5093-4262-81c5-edddad573201",
  :id "3124d886-7a5d-4c1e-9fc3-2946b1b3cfaa"}
 {:ns "user",
  :value "499999500000",
  :session "2ba81681-5093-4262-81c5-edddad573201",
  :id "3124d886-7a5d-4c1e-9fc3-2946b1b3cfaa"}
 {:status ["done"],
  :session "2ba81681-5093-4262-81c5-edddad573201",
  :id "3124d886-7a5d-4c1e-9fc3-2946b1b3cfaa"})

The different message types and their schemas are detailed here.

Embedding nREPL, starting a server

If your project uses Leiningen (v2 or higher), you already have access to an nREPL server for your project via lein repl.

Otherwise, it can be extremely useful to have your application host a REPL server whereever it might be deployed; this can greatly simplify debugging, sanity-checking, and so on.

nREPL provides a socket-based server that you can trivially start from your application. Add it to your project's dependencies, and add code like this to your app:

=> (use '[clojure.tools.nrepl.server :only (start-server stop-server)])
=> (defonce server (start-server :port 7888))

Depending on what the lifecycle of your application is, whether you want to be able to easily restart the server, etc., you might want to put the value start-server returns into an atom or somesuch. Anyway, once your app is running an nREPL server, you can connect to it from a tool like Leiningen or Counterclockwise or Reply, or from another Clojure process:

=> (with-open [conn (repl/connect :port 7888)]
     (-> (repl/client conn 1000)
       (repl/message {:op :eval :code "(+ 1 1)"})

You can stop the server with (stop-server server).

Server options

  • tty
  • http
  • implementing your own transport / handler(s)

Building nREPL

Releases are available from Maven Central, and SNAPSHOT builds from master's HEAD are automatically deployed to Sonatype's OSS repository (see this for how to configure Leiningen or Maven to use OSS-snapshots), so building nREPL shouldn't ever be necessary. That said:

  1. Clone the repo
  2. Make sure you have maven installed
  3. Run the maven build; run either:
    1. mvn package: This will produce an nREPL jar file in the target directory, and run all tests against Clojure 1.2.0.
    2. mvn verify: This does the same, but also runs the tests with other Clojure "profiles" (currently v1.1.0 and v1.1.0 + clojure-contrib).

Need Help?

Ping cemerick on freenode irc or twitter.

Why nREPL?

nREPL has been designed with the aim of ensuring that it satisfies the requirements of both application developers (in support of activities ranging from interactive remote debugging and experimentation in development contexts through to more advanced use cases such as updating deployed applications) as well as toolmakers (providing a standard way to connect to and introspect running environments as a way of informing user interfaces of all kinds, including "standard" interactive, text-based REPLs).

The network protocol used is simple, depending neither on JVM or Clojure specifics, thereby allowing (encouraging?) the development of non-Clojure REPL clients. The REPLs operational semantics are such that essentially any future non-JVM Clojure implementations should be able to implement it, with allowances for hosts that lack the concurrency primitives to support e.g. asynchronous evaluation, interrupts, etc.

For more information about the motivation, architecture, use cases, and discussion related to nREPL, see the see the original design notes, available here, and the notes and discussion around its recent redesign.


nREPL largely consists of three abstractions: handlers, middleware, and transports. These are largely analogous to the handlers, middleware, and adapters of Ring, though there are some important semantic differences, especially around transports.


nREPL is fundamentally message-oriented, where a message is a map of key/value pairs. Handlers are functions accept a single incoming message as an argument, and should return a truthy value indicating whether or not the provided message was processed. An nREPL server is started with a single handler function, which will be used to process messages for the lifetime of the server.

(This is because the most prevalent operation — evaluation of Clojure code — is fundamentally asynchronous.)

Messages are guaranteed to contain the following slots:

  • :transport The transport that should be used to send all responses precipitated by a given message.
  • :op The operation to perform; roughly, the type of message

Depending on its :op, a message might be required to contain other slots, and might optionally contain others. Each request should contain a unique :id.

Responses are also messages, maps of key/value pairs, the content of which depend entirely upon the type of message for which the responses are produced. Every request must provoke at least one and potentially many response messages, each of which should contain an :id slot echoing that of the provoking request. Once a handler has completely processed a message, a response containing a :status of :done must be sent. Some operations necessitate that additional responses related to the processing of a request are sent after a :done :status is reported (e.g. delivering content written to *out* by evaluated code that started a future).

Other statuses are possible, depending upon the semantics of the :op being handled; in particular, if the message is malformed or incomplete for a particular :op, then a response with an :error :status should be sent, potentially with additional information about the nature of the problem.

It is possible for an nREPL server to send messages to a client that are not a direct response to a request (e.g. streaming content written to System/out might be started/stopped by requests, but messages containing such content can't be considered responses to those requests).

If the handler being used by an nREPL server returns a logically false value (indicating that a message's :op was unrecognized), then the the server will respond with a message containing a :status of "unknown-op".

Generally, the handler that is provided as the :handler to clojure.tools.nrepl.server/start-server is built up as a result of composing multiple pieces of middleware.


Middleware are higher-order functions that compose additional functionality onto or around a handler. For example, some middleware that handles a "time?" :op by replying with the local time on the server:

(require '[clojure.tools.nrepl.transport :as t])
(use '[clojure.tools.nrepl.misc :only (response-for)])

(defn current-time
  (fn [{:keys [op transport] :as msg}]
    (if (= "time?" op)
      (t/send transport (response-for msg :status :done :time (System/currentTimeMillis)))
      (h msg))))

A little silly, perhaps, but you should get the idea. Nearly all of the same patterns and expectations associated with Ring middleware should be applicable to nREPL middleware.

It is recommended that (constantly false) always be the base handler; this ensures that unhandled messages will always yield a logically false return value. For example, nREPL's default handler is constructed like so in the clojure.tools.nrepl.server namespace:

(defn default-handler
  "A default handler supporting interruptible evaluation, stdin, sessions, and
   readable representations of evaluated expressions via `pr`."
  (-> (constantly false)

This combination — when paired with a suitably-capable client — aims to match and exceed the functionality offered by the standard Clojure REPL. Please see the documentation for each of those middleware functions for details as to what they expect in requests, and what they emit for responses.


Transports are roughly analogous to Ring's adapters: they provide an implementation of a common protocol (clojure.tools.nrepl.transport.Transport) to enable nREPL to send and receive messages without regard for the underlying mode of communication. Some transport implementations may be usable by a client as well as the server, but this is not expected to be common.

nREPL's default transport utilizes bencode-encoded messages sent over sockets; it is used by default by clojure.tools.nrepl.server/start-server and clojure.tools.nrepl/connect.

On the other hand, a Transport implementation suitable for exposing nREPL over plain-text sockets (clojure.tools.nrepl.transport/tty) is only usable by the server. Its only purpose is to enable the use of simpler/less capable tools (e.g. telnet et al.) to connect to an nREPL backend, so there's little reason to support client usage in such a scenario. Simiarly, an HTTP Transport (taking the form of a Ring handler) would expect clients to connect via e.g. curl, a browser, or a Javascript-powered HTTP console.

Change Log


Top-to-bottom redesign


Never released; initial prototype of "rich content" support that (in part) helped motivate a re-examination of the underlying protocol and design.


  • added Clojure 1.3.0 (ALPHA) compatibility


  • fixed (hacked) obtaining clojure.test output when clojure.test is initially loaded within an nREPL session
  • eliminated 1-minute default timeout on expression evaluation
  • all standard REPL var bindings are now properly established and maintained within a session


Thanks to the following Clojure masters for their helpful feedback during the initial design phases of nREPL:

  • Justin Balthrop
  • Meikel Brandmeyer
  • Hugo Duncan
  • Christophe Grand
  • Anthony Grimes
  • Phil Hagelberg
  • Rich Hickey
  • Chris Houser
  • Colin Jones
  • Laurent Petit
  • Eric Thorsen


Copyright © 2010 - 2012 Chas Emerick and contributors.

Licensed under the EPL. (See the file epl.html.)