diff --git a/DESCRIPTION b/DESCRIPTION
index 03c3b4bef..7dfe9b62c 100644
--- a/DESCRIPTION
+++ b/DESCRIPTION
@@ -107,6 +107,8 @@ Suggests:
   styler,
   visNetwork,
   webshot
+Remotes:
+  mschubert/clustermq@develop
 VignetteBuilder: knitr
 Encoding: UTF-8
 RoxygenNote: 6.1.0
diff --git a/NAMESPACE b/NAMESPACE
index acb4e0712..3ab4c2713 100644
--- a/NAMESPACE
+++ b/NAMESPACE
@@ -26,6 +26,7 @@ export(check)
 export(check_plan)
 export(clean)
 export(cleaned_namespaces)
+export(cmq_build)
 export(cmq_staged_build)
 export(code_to_plan)
 export(config)
diff --git a/NEWS.md b/NEWS.md
index 530020f1c..028026b6c 100644
--- a/NEWS.md
+++ b/NEWS.md
@@ -1,5 +1,6 @@
 # Version 5.4.0.9000
 
+- Add a proper [`clustermq`](https://github.com/mschubert/clustermq)-based parallel backend: `make(parallelism = "clustermq")`.
 - Smooth the edges in `vis_drake_graph()` and `render_drake_graph()`.
 - Make hover text slightly more readable in in `vis_drake_graph()` and `render_drake_graph()`.
 - Align hover text properly in `vis_drake_graph()` using the "title" node column.
diff --git a/R/clustermq.R b/R/clustermq.R
new file mode 100644
index 000000000..2759a1274
--- /dev/null
+++ b/R/clustermq.R
@@ -0,0 +1,145 @@
+run_clustermq <- function(config){
+  assert_pkg("clustermq", version = "0.8.4.99")
+  config$queue <- new_priority_queue(config = config, jobs = 1)
+  if (!config$queue$empty()){
+    config$workers <- clustermq::workers(
+      n_jobs = config$jobs,
+      template = config$template
+    )
+    cmq_set_common_data(config)
+    config$counter <- new.env(parent = emptyenv())
+    config$counter$remaining <- config$queue$size()
+    cmq_master(config)
+  }
+}
+
+cmq_set_common_data <- function(config){
+  export <- list()
+  if (identical(config$envir, globalenv())){
+    export <- as.list(config$envir, all.names = TRUE) # nocov
+  }
+  export$config <- config
+  config$workers$set_common_data(
+    export = export,
+    fun = identity,
+    const = list(),
+    rettype = list(),
+    common_seed = config$seed,
+    token = "set_common_data_token"
+  )
+}
+
+cmq_master <- function(config){
+  on.exit(config$workers$finalize())
+  while (config$counter$remaining > 0){
+    msg <- config$workers$receive_data()
+    cmq_conclude_build(msg = msg, config = config)
+    if (!identical(msg$token, "set_common_data_token")){
+      config$workers$send_common_data()
+    } else if (!config$queue$empty()){
+      cmq_send_target(config)
+    } else {
+      config$workers$send_shutdown_worker()
+    }
+  }
+  if (config$workers$cleanup()){
+    on.exit()
+  }
+}
+
+cmq_send_target <- function(config){
+  target <- config$queue$pop0()
+  # Longer tests will catch this:
+  if (!length(target)){
+    config$workers$send_wait() # nocov
+    return() # nocov
+  }
+  meta <- drake_meta(target = target, config = config)
+  # Target should not even be in the priority queue
+  # nocov start
+  if (!should_build_target(target = target, meta = meta, config = config)){
+    console_skip(target = target, config = config)
+    cmq_conclude_target(target = target, config = config)
+    config$workers$send_wait()
+    return()
+  }
+  # nocov end
+  meta$start <- proc.time()
+  announce_build(target = target, meta = meta, config = config)
+  prune_envir(targets = target, config = config, jobs = 1)
+  deps <- cmq_deps_list(target = target, config = config)
+  config$workers$send_call(
+    expr = drake::cmq_build(
+      target = target,
+      meta = meta,
+      deps = deps,
+      config = config
+    ),
+    env = list(target = target, meta = meta, deps = deps)
+  )
+}
+
+cmq_deps_list <- function(target, config){
+  deps <- dependencies(targets = target, config = config) %>%
+    intersect(config$plan$target)
+  lapply(
+    X = deps,
+    FUN = function(name){
+      config$envir[[name]]
+    }
+  ) %>%
+    setNames(deps)
+}
+
+#' @title Build a target using the clustermq backend
+#' @description For internal use only
+#' @export
+#' @keywords internal
+#' @inheritParams drake_build
+#' @param target target name
+#' @param meta list of metadata
+#' @param deps named list of target dependencies
+#' @param config a [drake_config()] list
+cmq_build <- function(target, meta, deps, config){
+  if (identical(config$garbage_collection, TRUE)){
+    gc()
+  }
+  do_prework(config = config, verbose_packages = FALSE)
+  for (dep in names(deps)){
+    config$envir[[dep]] <- deps[[dep]]
+  }
+  build <- just_build(target = target, meta = meta, config = config)
+  build$checksum <- mc_output_file_checksum(target, config)
+  build
+}
+
+cmq_conclude_build <- function(msg, config){
+  build <- msg$result
+  if (is.null(build)){
+    return()
+  }
+  cmq_conclude_target(target = build$target, config = config)
+  set_attempt_flag(key = build$target, config = config)
+  mc_wait_outfile_checksum(
+    target = build$target,
+    checksum = build$checksum,
+    config = config
+  )
+  conclude_build(
+    target = build$target,
+    value = build$value,
+    meta = build$meta,
+    config = config
+  )
+}
+
+cmq_conclude_target <- function(target, config){
+  revdeps <- dependencies(
+    targets = target,
+    config = config,
+    reverse = TRUE
+  ) %>%
+    intersect(y = config$queue$list())
+  config$queue$decrease_key(targets = revdeps)
+  config$counter$remaining <- config$counter$remaining - 1
+}
diff --git a/R/mclapply.R b/R/mclapply.R
index e22e42ffc..9e77743a5 100644
--- a/R/mclapply.R
+++ b/R/mclapply.R
@@ -4,7 +4,7 @@ run_mclapply <- function(config){
   }
   assert_pkg("txtq")
   mc_init_worker_cache(config)
-  tmp <- mclapply(
+  parallel::mclapply(
     X = mc_worker_id(c(0, seq_len(config$jobs))),
     FUN = mc_process,
     mc.cores = config$jobs + 1,
diff --git a/R/parallel.R b/R/parallel.R
index eaab08e05..0f95448b2 100644
--- a/R/parallel.R
+++ b/R/parallel.R
@@ -16,7 +16,7 @@ lightly_parallelize <- function(X, FUN, jobs = 1, ...) {
   if (is.atomic(X)){
     lightly_parallelize_atomic(X = X, FUN = FUN, jobs = jobs, ...)
   } else {
-    mclapply(X = X, FUN = FUN, mc.cores = jobs, ...)
+    safe_mclapply(X = X, FUN = FUN, mc.cores = jobs, ...)
   }
 }
 
@@ -24,10 +24,20 @@ lightly_parallelize_atomic <- function(X, FUN, jobs = 1, ...){
   jobs <- safe_jobs_imports(jobs)
   keys <- unique(X)
   index <- match(X, keys)
-  values <- mclapply(X = keys, FUN = FUN, mc.cores = jobs, ...)
+  values <- safe_mclapply(X = keys, FUN = FUN, mc.cores = jobs, ...)
   values[index]
 }
 
+# Avoid SIGCHLD handler when mc.cores is 1.
+# Could help avoid zeromq interrupted system call errors.
+safe_mclapply <- function(X, FUN, mc.cores, ...){
+  if (mc.cores > 1){
+    parallel::mclapply(X = X, FUN = FUN, mc.cores = mc.cores, ...)
+  } else {
+    lapply(X = X, FUN = FUN, ...)
+  }
+}
+
 # x is parallelism or jobs
 imports_setting <- function(x){
   if (length(x) < 2){
diff --git a/R/parallel_ui.R b/R/parallel_ui.R
index b0b3ff2b9..985d712f6 100644
--- a/R/parallel_ui.R
+++ b/R/parallel_ui.R
@@ -105,6 +105,7 @@ parallelism_choices <- function(distributed_only = FALSE) {
     "parLapply_staged"
   )
   distributed <- c(
+    "clustermq",
     "clustermq_staged",
     "future",
     "future_lapply",
diff --git a/R/priority_queue.R b/R/priority_queue.R
index d7e85e184..4427cf758 100644
--- a/R/priority_queue.R
+++ b/R/priority_queue.R
@@ -1,4 +1,4 @@
-new_priority_queue <- function(config){
+new_priority_queue <- function(config, jobs = config$jobs_imports){
   config$graph <- config$schedule
   targets <- V(config$graph)$name
   if (!length(targets)){
@@ -9,7 +9,7 @@ new_priority_queue <- function(config){
     FUN = function(target){
       length(dependencies(targets = target, config = config))
     },
-    jobs = config$jobs_imports
+    jobs = jobs
   ) %>%
     unlist
   priorities <- rep(Inf, length(targets))
diff --git a/R/staged.R b/R/staged.R
index 37931a8b0..f31b23479 100644
--- a/R/staged.R
+++ b/R/staged.R
@@ -1,4 +1,4 @@
-next_stage <- function(config, schedule) {
+next_stage <- function(config, schedule, jobs) {
   targets <- character(0)
   old_leaves <- NULL
   meta_list <- list()
@@ -14,7 +14,7 @@ next_stage <- function(config, schedule) {
     new_meta <- lightly_parallelize(
       X = new_leaves,
       FUN = drake_meta,
-      jobs = config$jobs_imports,
+      jobs = jobs,
       config = config
     )
     names(new_meta) <- new_leaves
@@ -27,7 +27,7 @@ next_stage <- function(config, schedule) {
           config = config
         )
       },
-      jobs = config$jobs_imports
+      jobs = jobs
     ) %>%
       unlist
     targets <- c(targets, new_leaves[do_build])
@@ -47,7 +47,11 @@ run_mclapply_staged <- function(config){
   config$jobs <- safe_jobs(config$jobs)
   schedule <- config$schedule
   while (length(V(schedule)$name)){
-    stage <- next_stage(config = config, schedule = schedule)
+    stage <- next_stage(
+      config = config,
+      schedule = schedule,
+      jobs = config$jobs
+    )
     schedule <- stage$schedule
     if (!length(stage$targets)){
       break
@@ -55,7 +59,7 @@ run_mclapply_staged <- function(config){
       set_attempt_flag(key = "_attempt", config = config)
     }
     prune_envir(targets = stage$targets, config = config, jobs = config$jobs)
-    tmp <- mclapply(
+    parallel::mclapply(
       X = stage$targets,
       FUN = function(target){
         build_and_store(
@@ -100,7 +104,11 @@ run_parLapply_staged <- function(config) { # nolint
   )
   schedule <- config$schedule
   while (length(V(schedule)$name)){
-    stage <- next_stage(config = config, schedule = schedule)
+    stage <- next_stage(
+      config = config,
+      schedule = schedule,
+      jobs = config$jobs
+    )
     schedule <- stage$schedule
     if (!length(stage$targets)){
       break
@@ -146,7 +154,11 @@ run_future_lapply_staged <- function(config){
   prepare_distributed(config = config)
   schedule <- config$schedule
   while (length(V(schedule)$name)){
-    stage <- next_stage(config = config, schedule = schedule)
+    stage <- next_stage(
+      config = config,
+      schedule = schedule,
+      jobs = config$jobs_imports
+    )
     schedule <- stage$schedule
     if (!length(stage$targets)){
       # Keep in case outdated targets are ever back in the schedule.
@@ -165,26 +177,33 @@ run_future_lapply_staged <- function(config){
 }
 
 run_clustermq_staged <- function(config){
-  assert_pkg("clustermq")
+  assert_pkg("clustermq", version = "0.8.4.99")
   schedule <- config$schedule
-  workers <- clustermq::workers(
-    n_jobs = config$jobs,
-    template = config$template
-  )
-  on.exit(workers$finalize())
+  workers <- NULL
   while (length(V(schedule)$name)){
-    stage <- next_stage(config = config, schedule = schedule)
+    stage <- next_stage(
+      config = config,
+      schedule = schedule,
+      jobs = 1 # config$jobs # nolint
+    )
     schedule <- stage$schedule
     if (!length(stage$targets)){
       # Keep in case outdated targets are ever back in the schedule.
       break # nocov
-    } else if (any(stage$targets %in% config$plan$target)){
+    } else if (is.null(workers)){
+      workers <- clustermq::workers(
+        n_jobs = config$jobs,
+        template = config$template
+      )
+      on.exit(workers$finalize())
+    }
+    if (any(stage$targets %in% config$plan$target)){
       set_attempt_flag(key = "_attempt", config = config)
     }
     prune_envir(
       targets = stage$targets,
       config = config,
-      jobs = config$jobs_imports
+      jobs = 1 # config$jobs_imports # nolint
     )
     export <- list()
     if (identical(config$envir, globalenv())){
@@ -202,7 +221,7 @@ run_clustermq_staged <- function(config){
           config = config
         )
       },
-      jobs = config$jobs_imports
+      jobs = 1 # config$jobs_imports # nolint
     )
     builds <- clustermq::Q(
       stage$targets,
@@ -234,13 +253,16 @@ run_clustermq_staged <- function(config){
           config = config
         )
       },
-      jobs = config$jobs_imports
+      jobs = 1 # config$jobs_imports # nolint
     )
   }
+  if (!is.null(workers) && workers$cleanup()){
+    on.exit()
+  }
   invisible()
 }
 
-#' @title Build a target using the clustermq backend
+#' @title Build a target using the clustermq_staged backend
 #' @description For internal use only
 #' @export
 #' @keywords internal
diff --git a/inst/testing/scenarios.csv b/inst/testing/scenarios.csv
index f60d66a1f..84706c2e5 100644
--- a/inst/testing/scenarios.csv
+++ b/inst/testing/scenarios.csv
@@ -5,6 +5,7 @@ envir,parallelism,jobs,backend,caching
 "local","mclapply",1,,
 "local","mclapply",9,,
 "local","mclapply_staged",2,,
+"local","clustermq",2,,
 "local","clustermq_staged",9,,
 "local","Makefile",9,,
 "local","future_lapply",2,"future::multisession",
@@ -16,6 +17,7 @@ envir,parallelism,jobs,backend,caching
 "global","mclapply",1,,
 "global","mclapply",2,,
 "global","mclapply_staged",9,,
+"global","clustermq",9,,
 "global","clustermq_staged",2,,
 "global","Makefile",2,,
 "global","future_lapply",1,"future::multisession",
diff --git a/man/cmq_build.Rd b/man/cmq_build.Rd
new file mode 100644
index 000000000..ce69c8bdf
--- /dev/null
+++ b/man/cmq_build.Rd
@@ -0,0 +1,21 @@
+% Generated by roxygen2: do not edit by hand
+% Please edit documentation in R/clustermq.R
+\name{cmq_build}
+\alias{cmq_build}
+\title{Build a target using the clustermq backend}
+\usage{
+cmq_build(target, meta, deps, config)
+}
+\arguments{
+\item{target}{target name}
+
+\item{meta}{list of metadata}
+
+\item{deps}{named list of target dependencies}
+
+\item{config}{a \code{\link[=drake_config]{drake_config()}} list}
+}
+\description{
+For internal use only
+}
+\keyword{internal}
diff --git a/man/cmq_staged_build.Rd b/man/cmq_staged_build.Rd
index 09cd93cba..a59de5df3 100644
--- a/man/cmq_staged_build.Rd
+++ b/man/cmq_staged_build.Rd
@@ -2,7 +2,7 @@
 % Please edit documentation in R/staged.R
 \name{cmq_staged_build}
 \alias{cmq_staged_build}
-\title{Build a target using the clustermq backend}
+\title{Build a target using the clustermq_staged backend}
 \usage{
 cmq_staged_build(target, meta_list, config)
 }
diff --git a/tests/testthat/test-clustermq.R b/tests/testthat/test-clustermq.R
index 8c9ca535c..c46e0eca6 100644
--- a/tests/testthat/test-clustermq.R
+++ b/tests/testthat/test-clustermq.R
@@ -1,6 +1,6 @@
 drake_context("clustermq")
 
-test_with_dir("clustermq_staged parallelism", {
+test_with_dir("clustermq parallelism", {
   skip_on_cran()
   if ("package:clustermq" %in% search()){
     eval(parse(text = "detach('package:clustermq', unload = TRUE)"))
@@ -10,25 +10,29 @@ test_with_dir("clustermq_staged parallelism", {
   skip_on_os("windows")
   scenario <- get_testing_scenario()
   e <- eval(parse(text = scenario$envir))
-  jobs <- scenario$jobs
+  jobs <- scenario$jobs # ignoring for now, using 2 jobs
   load_mtcars_example(envir = e)
-  make(
-    e$my_plan,
-    parallelism = "clustermq_staged",
-    jobs = jobs,
-    envir = e,
-    verbose = 4
-  )
-  config <- drake_config(e$my_plan, envir = e)
-  expect_equal(outdated(config), character(0))
-  make(
-    e$my_plan,
-    parallelism = "clustermq_staged",
-    jobs = jobs,
-    envir = e,
-    verbose = 4
-  )
-  expect_equal(justbuilt(config), character(0))
+  for (parallelism in c("clustermq", "clustermq_staged")){
+    make(
+      e$my_plan,
+      parallelism = parallelism,
+      jobs = 2,
+      envir = e,
+      verbose = 4,
+      garbage_collection = TRUE
+    )
+    config <- drake_config(e$my_plan, envir = e)
+    expect_equal(outdated(config), character(0))
+    make(
+      e$my_plan,
+      parallelism = parallelism,
+      jobs = 2,
+      envir = e,
+      verbose = 4
+    )
+    expect_equal(justbuilt(config), character(0))
+    clean(destroy = TRUE)
+  }
   if ("package:clustermq" %in% search()){
     eval(parse(text = "detach('package:clustermq', unload = TRUE)"))
   }