diff --git a/CHANGELOG.md b/CHANGELOG.md
index d2dd760..c55e16b 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -11,6 +11,11 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 * Migrated codebase to V 0.3.2
 * Cron expression parser now uses bitfields instead of bool arrays
+* Added option to deploy using agent-server architecture instead of cron daemon
+* Allow force-building packages, meaning the build won't check if the
+  repository is already up to date
+* Allow scheduling builds on the server from the CLI tool instead of building
+  them locally
 
 ### Fixed
 
@@ -19,6 +24,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
     * CLI no longer exits with non-zero status code when removing/patching
       target
 * Allow NULL values for branch in database
+* Endpoint for adding targets now returns the correct id
 
 ## [0.4.0](https://git.rustybever.be/vieter-v/vieter/src/tag/0.4.0)
 
diff --git a/README.md b/README.md
index b9fff69..637d4c1 100644
--- a/README.md
+++ b/README.md
@@ -21,7 +21,8 @@ quicker.
 I chose [V](https://vlang.io/) as I've been very intrigued by this language for
 a while now. I wanted a fast language that I could code while relaxing, without
 having to exert too much mental effort & V seemed like the right choice for
-that.
+that. Sadly, this didn't quite turn out the way I expected, but I'm sticking
+with it anyways ;p
 
 ## Features
 
@@ -49,7 +50,7 @@ update`.
 
 I used to maintain a mirror that tracked the latest master, but nowadays, I
 maintain a Docker image containing the specific compiler version that Vieter
-builds with. Currently, this is V 0.3.
+builds with. Currently, this is V 0.3.2.
 
 ## Contributing
 
diff --git a/docs/content/configuration.md b/docs/content/configuration.md
index af941a2..95bf713 100644
--- a/docs/content/configuration.md
+++ b/docs/content/configuration.md
@@ -17,7 +17,7 @@ If a variable is both present in the config file & as an environment variable,
 the value in the environment variable is used.
 
 {{< hint info >}}
-**Note**  
+**Note**
 All environment variables can also be provided from a file by appending them
 with `_FILE`. This for example allows you to provide the API key from a Docker
 secrets file.
@@ -97,3 +97,25 @@ configuration variable required for each command.
   build`.
     * Default: `archlinux:base-devel`
 
+### `vieter agent`
+
+* `log_level`: log verbosity level. Value should be one of `FATAL`, `ERROR`,
+  `WARN`, `INFO` or `DEBUG`.
+    * Default: `WARN`
+* `address`: *public* URL of the Vieter repository server to build for. From
+  this server jobs are retrieved. All built packages are published to this
+  server.
+* `api_key`: API key of the above server.
+* `data_dir`: directory to store log file in.
+* `max_concurrent_builds`: how many builds to run at the same time.
+    * Default: `1`
+* `polling_frequency`: how often (in seconds) to poll the server for new
+  builds. Note that the agent might poll more frequently when it's actively
+  processing builds.
+* `image_rebuild_frequency`: Vieter periodically builds images that are then
+  used as a basis for running build containers. This is to prevent each build
+  from downloading an entire repository worth of dependencies. This setting
+  defines how frequently (in minutes) to rebuild these images.
+    * Default: `1440` (every 24 hours)
+* `arch`: architecture for which this agent should pull down builds (e.g.
+  `x86_64`)
diff --git a/docs/content/installation.md b/docs/content/installation.md
index 87b9cba..21eda64 100644
--- a/docs/content/installation.md
+++ b/docs/content/installation.md
@@ -21,7 +21,7 @@ branch. This branch will be the most up to date, but does not give any
 guarantees about stability, so beware!
 
 Thanks to the single-binary design of Vieter, this image can be used both for
-the repository server & the cron daemon.
+the repository server, the cron daemon and the agent.
 
 Below is an example compose file to set up both the repository server & the
 cron daemon:
@@ -76,7 +76,7 @@ architectures will build on both.
 ## Binary
 
 On the
-[releases](https://git.rustybever.be/vieter/vieter/releases)
+[releases](https://git.rustybever.be/vieter-v/vieter/releases)
 page, you can find statically compiled binaries for all
 released versions. This is the same binary as used inside
 the Docker images.
@@ -106,5 +106,5 @@ guarantee that a compiler update won't temporarily break them.
 
 ## Building from source
 
-The project [README](https://git.rustybever.be/vieter/vieter#building) contains
-instructions for building Vieter from source.
+The project [README](https://git.rustybever.be/vieter-v/vieter#building)
+contains instructions for building Vieter from source.
diff --git a/docs/content/usage/builds/schedule.md b/docs/content/usage/builds/schedule.md
index 38f76a4..de59e25 100644
--- a/docs/content/usage/builds/schedule.md
+++ b/docs/content/usage/builds/schedule.md
@@ -37,6 +37,6 @@ Each section can consist of as many of these parts as necessary.
 ## CLI tool
 
 The Vieter binary contains a command that shows you the next matching times for
-a given expression. This can be useful to understand the syntax. For more
+a given expression. This can be useful for understanding the syntax. For more
 information, see
 [vieter-schedule(1)](https://rustybever.be/man/vieter/vieter-schedule.1.html).
diff --git a/src/agent/agent.v b/src/agent/agent.v
new file mode 100644
index 0000000..69b9947
--- /dev/null
+++ b/src/agent/agent.v
@@ -0,0 +1,27 @@
+module agent
+
+import log
+import os
+import util
+
+const log_file_name = 'vieter.agent.log'
+
+// agent starts an agent service
+pub fn agent(conf Config) ! {
+	log_level := log.level_from_tag(conf.log_level) or {
+		return error('Invalid log level. The allowed values are FATAL, ERROR, WARN, INFO & DEBUG.')
+	}
+
+	mut logger := log.Log{
+		level: log_level
+	}
+
+	os.mkdir_all(conf.data_dir) or { util.exit_with_message(1, 'Failed to create data directory.') }
+
+	log_file := os.join_path_single(conf.data_dir, agent.log_file_name)
+	logger.set_full_logpath(log_file)
+	logger.log_to_console_too()
+
+	mut d := agent_init(logger, conf)
+	d.run()
+}
diff --git a/src/agent/cli.v b/src/agent/cli.v
new file mode 100644
index 0000000..1535e17
--- /dev/null
+++ b/src/agent/cli.v
@@ -0,0 +1,31 @@
+module agent
+
+import cli
+import conf as vconf
+
+struct Config {
+pub:
+	log_level string = 'WARN'
+	// Architecture that the agent represents
+	arch                    string
+	api_key                 string
+	address                 string
+	data_dir                string
+	max_concurrent_builds   int = 1
+	polling_frequency       int = 30
+	image_rebuild_frequency int = 1440
+}
+
+// cmd returns the cli module that handles the cron daemon.
+pub fn cmd() cli.Command {
+	return cli.Command{
+		name: 'agent'
+		description: 'Start an agent daemon.'
+		execute: fn (cmd cli.Command) ! {
+			config_file := cmd.flags.get_string('config-file')!
+			conf := vconf.load<Config>(prefix: 'VIETER_', default_path: config_file)!
+
+			agent(conf)!
+		}
+	}
+}
diff --git a/src/agent/daemon.v b/src/agent/daemon.v
new file mode 100644
index 0000000..0647733
--- /dev/null
+++ b/src/agent/daemon.v
@@ -0,0 +1,178 @@
+module agent
+
+import log
+import sync.stdatomic
+import build { BuildConfig }
+import client
+import time
+import os
+
+const (
+	build_empty   = 0
+	build_running = 1
+	build_done    = 2
+)
+
+struct AgentDaemon {
+	logger shared log.Log
+	conf   Config
+	client client.Client
+mut:
+	images ImageManager
+	// Which builds are currently running; length is conf.max_concurrent_builds
+	builds []BuildConfig
+	// Atomic variables used to detect when a build has finished; length is
+	// conf.max_concurrent_builds
+	atomics []u64
+}
+
+// agent_init initializes a new agent
+fn agent_init(logger log.Log, conf Config) AgentDaemon {
+	mut d := AgentDaemon{
+		logger: logger
+		client: client.new(conf.address, conf.api_key)
+		conf: conf
+		images: new_image_manager(conf.image_rebuild_frequency * 60)
+		builds: []BuildConfig{len: conf.max_concurrent_builds}
+		atomics: []u64{len: conf.max_concurrent_builds}
+	}
+
+	return d
+}
+
+// run starts the actual agent daemon. This function will run forever.
+pub fn (mut d AgentDaemon) run() {
+	// This is just so that the very first time the loop is ran, the jobs are
+	// always polled
+	mut last_poll_time := time.now().add_seconds(-d.conf.polling_frequency)
+	mut sleep_time := 1 * time.second
+	mut finished, mut empty := 0, 0
+
+	for {
+		finished, empty = d.update_atomics()
+
+		// No new finished builds and no free slots, so there's nothing to be
+		// done
+		if finished + empty == 0 {
+			time.sleep(1 * time.second)
+			continue
+		}
+
+		// Builds have finished, so old builder images might have freed up.
+		// TODO this might query the docker daemon too frequently.
+		if finished > 0 {
+			d.images.clean_old_images()
+		}
+
+		// The agent will always poll for new jobs after at most
+		// `polling_frequency` seconds. However, when jobs have finished, the
+		// agent will also poll for new jobs. This is because jobs are often
+		// clustered together (especially when mostly using the global cron
+		// schedule), so there's a much higher chance jobs are available.
+		if finished > 0 || time.now() >= last_poll_time.add_seconds(d.conf.polling_frequency) {
+			new_configs := d.client.poll_jobs(d.conf.arch, finished + empty) or {
+				d.lerror('Failed to poll jobs: $err.msg()')
+
+				// TODO pick a better delay here
+				time.sleep(5 * time.second)
+				continue
+			}
+			last_poll_time = time.now()
+
+			for config in new_configs {
+				// TODO handle this better than to just skip the config
+				// Make sure a recent build base image is available for
+				// building the config
+				d.images.refresh_image(config.base_image) or {
+					d.lerror(err.msg())
+					continue
+				}
+				d.start_build(config)
+			}
+
+			// No new jobs were scheduled and the agent isn't doing anything,
+			// so we just wait until the next polling period.
+			if new_configs.len == 0 && finished + empty == d.conf.max_concurrent_builds {
+				sleep_time = time.now() - last_poll_time
+			}
+		}
+		// The agent is not doing anything, so we just wait until the next poll
+		// time
+		else if finished + empty == d.conf.max_concurrent_builds {
+			sleep_time = time.now() - last_poll_time
+		}
+
+		time.sleep(sleep_time)
+	}
+}
+
+// update_atomics checks for each build whether it's completed, and sets it to
+// empty again if so. The return value is a tuple `(finished, empty)` where
+// `finished` is how many builds were just finished and thus set to empty, and
+// `empty` is how many build slots were already empty. The amount of running
+// builds can then be calculated by substracting these two values from the
+// total allowed concurrent builds.
+fn (mut d AgentDaemon) update_atomics() (int, int) {
+	mut finished := 0
+	mut empty := 0
+
+	for i in 0 .. d.atomics.len {
+		if stdatomic.load_u64(&d.atomics[i]) == agent.build_done {
+			stdatomic.store_u64(&d.atomics[i], agent.build_empty)
+			finished++
+		} else if stdatomic.load_u64(&d.atomics[i]) == agent.build_empty {
+			empty++
+		}
+	}
+
+	return finished, empty
+}
+
+// start_build starts a build for the given BuildConfig.
+fn (mut d AgentDaemon) start_build(config BuildConfig) bool {
+	for i in 0 .. d.atomics.len {
+		if stdatomic.load_u64(&d.atomics[i]) == agent.build_empty {
+			stdatomic.store_u64(&d.atomics[i], agent.build_running)
+			d.builds[i] = config
+
+			go d.run_build(i, config)
+
+			return true
+		}
+	}
+
+	return false
+}
+
+// run_build actually starts the build process for a given target.
+fn (mut d AgentDaemon) run_build(build_index int, config BuildConfig) {
+	d.linfo('started build: $config')
+
+	// 0 means success, 1 means failure
+	mut status := 0
+
+	new_config := BuildConfig{
+		...config
+		base_image: d.images.get(config.base_image)
+	}
+
+	res := build.build_config(d.client.address, d.client.api_key, new_config) or {
+		d.ldebug('build_config error: $err.msg()')
+		status = 1
+
+		build.BuildResult{}
+	}
+
+	if status == 0 {
+		d.linfo('Uploading build logs for $config')
+
+		// TODO use the arch value here
+		build_arch := os.uname().machine
+		d.client.add_build_log(config.target_id, res.start_time, res.end_time, build_arch,
+			res.exit_code, res.logs) or { d.lerror('Failed to upload logs for $config') }
+	} else {
+		d.lwarn('an error occurred during build: $config')
+	}
+
+	stdatomic.store_u64(&d.atomics[build_index], agent.build_done)
+}
diff --git a/src/agent/images.v b/src/agent/images.v
new file mode 100644
index 0000000..185192e
--- /dev/null
+++ b/src/agent/images.v
@@ -0,0 +1,79 @@
+module agent
+
+import time
+import docker
+import build
+
+// An ImageManager is a utility that creates builder images from given base
+// images, updating these builder images if they've become too old. This
+// structure can manage images from any number of base images, paving the way
+// for configurable base images per target/repository.
+struct ImageManager {
+	max_image_age int [required]
+mut:
+	// For each base image, one or more builder images can exist at the same
+	// time
+	images map[string][]string [required]
+	// For each base image, we track when its newest image was built
+	timestamps map[string]time.Time [required]
+}
+
+// new_image_manager initializes a new image manager.
+fn new_image_manager(max_image_age int) ImageManager {
+	return ImageManager{
+		max_image_age: max_image_age
+		images: map[string][]string{}
+		timestamps: map[string]time.Time{}
+	}
+}
+
+// get returns the name of the newest image for the given base image. Note that
+// this function should only be called *after* a first call to `refresh_image`.
+pub fn (m &ImageManager) get(base_image string) string {
+	return m.images[base_image].last()
+}
+
+// refresh_image builds a new builder image from the given base image if the
+// previous builder image is too old or non-existent. This function will do
+// nothing if these conditions aren't met, so it's safe to call it every time
+// you want to ensure an image is up to date.
+fn (mut m ImageManager) refresh_image(base_image string) ! {
+	if base_image in m.timestamps
+		&& m.timestamps[base_image].add_seconds(m.max_image_age) > time.now() {
+		return
+	}
+
+	// TODO use better image tags for built images
+	new_image := build.create_build_image(base_image) or {
+		return error('Failed to build builder image from base image $base_image')
+	}
+
+	m.images[base_image] << new_image
+	m.timestamps[base_image] = time.now()
+}
+
+// clean_old_images removes all older builder images that are no longer in use.
+// The function will always leave at least one builder image, namely the newest
+// one.
+fn (mut m ImageManager) clean_old_images() {
+	mut dd := docker.new_conn() or { return }
+
+	defer {
+		dd.close() or {}
+	}
+
+	mut i := 0
+
+	for image in m.images.keys() {
+		i = 0
+
+		for i < m.images[image].len - 1 {
+			// For each builder image, we try to remove it by calling the Docker
+			// API. If the function returns an error or false, that means the image
+			// wasn't deleted. Therefore, we move the index over. If the function
+			// returns true, the array's length has decreased by one so we don't
+			// move the index.
+			dd.remove_image(m.images[image][i]) or { i += 1 }
+		}
+	}
+}
diff --git a/src/agent/log.v b/src/agent/log.v
new file mode 100644
index 0000000..cd59207
--- /dev/null
+++ b/src/agent/log.v
@@ -0,0 +1,35 @@
+module agent
+
+import log
+
+// log a message with the given level
+pub fn (mut d AgentDaemon) log(msg string, level log.Level) {
+	lock d.logger {
+		d.logger.send_output(msg, level)
+	}
+}
+
+// lfatal create a log message with the fatal level
+pub fn (mut d AgentDaemon) lfatal(msg string) {
+	d.log(msg, log.Level.fatal)
+}
+
+// lerror create a log message with the error level
+pub fn (mut d AgentDaemon) lerror(msg string) {
+	d.log(msg, log.Level.error)
+}
+
+// lwarn create a log message with the warn level
+pub fn (mut d AgentDaemon) lwarn(msg string) {
+	d.log(msg, log.Level.warn)
+}
+
+// linfo create a log message with the info level
+pub fn (mut d AgentDaemon) linfo(msg string) {
+	d.log(msg, log.Level.info)
+}
+
+// ldebug create a log message with the debug level
+pub fn (mut d AgentDaemon) ldebug(msg string) {
+	d.log(msg, log.Level.debug)
+}
diff --git a/src/build/build.v b/src/build/build.v
index 247df6e..3d916bf 100644
--- a/src/build/build.v
+++ b/src/build/build.v
@@ -16,6 +16,22 @@ const (
 		'/usr/local/bin', '/usr/bin/site_perl', '/usr/bin/vendor_perl', '/usr/bin/core_perl']
 )
 
+pub struct BuildConfig {
+pub:
+	target_id  int
+	kind       string
+	url        string
+	branch     string
+	repo       string
+	base_image string
+	force      bool
+}
+
+// str return a single-line string representation of a build log
+pub fn (c BuildConfig) str() string {
+	return '{ target: $c.target_id, kind: $c.kind, url: $c.url, branch: $c.branch, repo: $c.repo, base_image: $c.base_image, force: $c.force }'
+}
+
 // create_build_image creates a builder image given some base image which can
 // then be used to build & package Arch images. It mostly just updates the
 // system, install some necessary packages & creates a non-root user to run
@@ -93,10 +109,25 @@ pub:
 	logs       string
 }
 
-// build_target builds, packages & publishes a given Arch package based on the
+// build_target builds the given target. Internally it calls `build_config`.
+pub fn build_target(address string, api_key string, base_image_id string, target &Target, force bool) !BuildResult {
+	config := BuildConfig{
+		target_id: target.id
+		kind: target.kind
+		url: target.url
+		branch: target.branch
+		repo: target.repo
+		base_image: base_image_id
+		force: force
+	}
+
+	return build_config(address, api_key, config)
+}
+
+// build_config builds, packages & publishes a given Arch package based on the
 // provided target. The base image ID should be of an image previously created
 // by create_build_image. It returns the logs of the container.
-pub fn build_target(address string, api_key string, base_image_id string, target &Target) !BuildResult {
+pub fn build_config(address string, api_key string, config BuildConfig) !BuildResult {
 	mut dd := docker.new_conn()!
 
 	defer {
@@ -104,14 +135,14 @@ pub fn build_target(address string, api_key string, base_image_id string, target
 	}
 
 	build_arch := os.uname().machine
-	build_script := create_build_script(address, target, build_arch)
+	build_script := create_build_script(address, config, build_arch)
 
 	// We convert the build script into a base64 string, which then gets passed
 	// to the container as an env var
 	base64_script := base64.encode_str(build_script)
 
 	c := docker.NewContainer{
-		image: '$base_image_id'
+		image: '$config.base_image'
 		env: [
 			'BUILD_SCRIPT=$base64_script',
 			'API_KEY=$api_key',
diff --git a/src/build/queue.v b/src/build/queue.v
new file mode 100644
index 0000000..7902173
--- /dev/null
+++ b/src/build/queue.v
@@ -0,0 +1,229 @@
+module build
+
+import models { Target }
+import cron.expression { CronExpression, parse_expression }
+import time
+import datatypes { MinHeap }
+import util
+
+struct BuildJob {
+pub mut:
+	// Time at which this build job was created/queued
+	created time.Time
+	// Next timestamp from which point this job is allowed to be executed
+	timestamp time.Time
+	// Required for calculating next timestamp after having pop'ed a job
+	ce CronExpression
+	// Actual build config sent to the agent
+	config BuildConfig
+	// Whether this is a one-time job
+	single bool
+}
+
+// Allows BuildJob structs to be sorted according to their timestamp in
+// MinHeaps
+fn (r1 BuildJob) < (r2 BuildJob) bool {
+	return r1.timestamp < r2.timestamp
+}
+
+// The build job queue is responsible for managing the list of scheduled builds
+// for each architecture. Agents receive jobs from this queue.
+pub struct BuildJobQueue {
+	// Schedule to use for targets without explicitely defined cron expression
+	default_schedule CronExpression
+	// Base image to use for targets without defined base image
+	default_base_image string
+mut:
+	mutex shared util.Dummy
+	// For each architecture, a priority queue is tracked
+	queues map[string]MinHeap<BuildJob>
+	// When a target is removed from the server or edited, its previous build
+	// configs will be invalid. This map allows for those to be simply skipped
+	// by ignoring any build configs created before this timestamp.
+	invalidated map[int]time.Time
+}
+
+// new_job_queue initializes a new job queue
+pub fn new_job_queue(default_schedule CronExpression, default_base_image string) BuildJobQueue {
+	return BuildJobQueue{
+		default_schedule: default_schedule
+		default_base_image: default_base_image
+		invalidated: map[int]time.Time{}
+	}
+}
+
+// insert_all executes insert for each architecture of the given Target.
+pub fn (mut q BuildJobQueue) insert_all(target Target) ! {
+	for arch in target.arch {
+		q.insert(target: target, arch: arch.value)!
+	}
+}
+
+[params]
+pub struct InsertConfig {
+	target Target [required]
+	arch   string [required]
+	single bool
+	force  bool
+	now    bool
+}
+
+// insert a new target's job into the queue for the given architecture. This
+// job will then be endlessly rescheduled after being pop'ed, unless removed
+// explicitely.
+pub fn (mut q BuildJobQueue) insert(input InsertConfig) ! {
+	lock q.mutex {
+		if input.arch !in q.queues {
+			q.queues[input.arch] = MinHeap<BuildJob>{}
+		}
+
+		mut job := BuildJob{
+			created: time.now()
+			single: input.single
+			config: BuildConfig{
+				target_id: input.target.id
+				kind: input.target.kind
+				url: input.target.url
+				branch: input.target.branch
+				repo: input.target.repo
+				// TODO make this configurable
+				base_image: q.default_base_image
+				force: input.force
+			}
+		}
+
+		if !input.now {
+			ce := if input.target.schedule != '' {
+				parse_expression(input.target.schedule) or {
+					return error("Error while parsing cron expression '$input.target.schedule' (id $input.target.id): $err.msg()")
+				}
+			} else {
+				q.default_schedule
+			}
+
+			job.timestamp = ce.next_from_now()!
+			job.ce = ce
+		} else {
+			job.timestamp = time.now()
+		}
+
+		q.queues[input.arch].insert(job)
+	}
+}
+
+// reschedule the given job by calculating the next timestamp and re-adding it
+// to its respective queue. This function is called by the pop functions
+// *after* having pop'ed the job.
+fn (mut q BuildJobQueue) reschedule(job BuildJob, arch string) ! {
+	new_timestamp := job.ce.next_from_now()!
+
+	new_job := BuildJob{
+		...job
+		created: time.now()
+		timestamp: new_timestamp
+	}
+
+	q.queues[arch].insert(new_job)
+}
+
+// pop_invalid pops all invalid jobs.
+fn (mut q BuildJobQueue) pop_invalid(arch string) {
+	for {
+		job := q.queues[arch].peek() or { return }
+
+		if job.config.target_id in q.invalidated
+			&& job.created < q.invalidated[job.config.target_id] {
+			// This pop *should* never fail according to the source code
+			q.queues[arch].pop() or {}
+		} else {
+			break
+		}
+	}
+}
+
+// peek shows the first job for the given architecture that's ready to be
+// executed, if present.
+pub fn (mut q BuildJobQueue) peek(arch string) ?BuildJob {
+	// Even peek requires a write lock, because pop_invalid can modify the data
+	// structure
+	lock q.mutex {
+		if arch !in q.queues {
+			return none
+		}
+
+		q.pop_invalid(arch)
+		job := q.queues[arch].peek()?
+
+		if job.timestamp < time.now() {
+			return job
+		}
+	}
+
+	return none
+}
+
+// pop removes the first job for the given architecture that's ready to be
+// executed from the queue and returns it, if present.
+pub fn (mut q BuildJobQueue) pop(arch string) ?BuildJob {
+	lock q.mutex {
+		if arch !in q.queues {
+			return none
+		}
+
+		q.pop_invalid(arch)
+		mut job := q.queues[arch].peek()?
+
+		if job.timestamp < time.now() {
+			job = q.queues[arch].pop()?
+
+			if !job.single {
+				// TODO how do we handle this properly? Is it even possible for a
+				// cron expression to not return a next time if it's already been
+				// used before?
+				q.reschedule(job, arch) or {}
+			}
+
+			return job
+		}
+	}
+
+	return none
+}
+
+// pop_n tries to pop at most n available jobs for the given architecture.
+pub fn (mut q BuildJobQueue) pop_n(arch string, n int) []BuildJob {
+	lock q.mutex {
+		if arch !in q.queues {
+			return []
+		}
+
+		mut out := []BuildJob{}
+
+		for out.len < n {
+			q.pop_invalid(arch)
+			mut job := q.queues[arch].peek() or { break }
+
+			if job.timestamp < time.now() {
+				job = q.queues[arch].pop() or { break }
+
+				if !job.single {
+					// TODO idem
+					q.reschedule(job, arch) or {}
+				}
+
+				out << job
+			} else {
+				break
+			}
+		}
+
+		return out
+	}
+
+	return []
+}
+
+// invalidate a target's old build jobs.
+pub fn (mut q BuildJobQueue) invalidate(target_id int) {
+	q.invalidated[target_id] = time.now()
+}
diff --git a/src/build/shell.v b/src/build/shell.v
index e573d53..ac61e07 100644
--- a/src/build/shell.v
+++ b/src/build/shell.v
@@ -1,7 +1,5 @@
 module build
 
-import models { Target }
-
 // escape_shell_string escapes any characters that could be interpreted
 // incorrectly by a shell. The resulting value should be safe to use inside an
 // echo statement.
@@ -23,13 +21,13 @@ pub fn echo_commands(cmds []string) []string {
 }
 
 // create_build_script generates a shell script that builds a given Target.
-fn create_build_script(address string, target &Target, build_arch string) string {
-	repo_url := '$address/$target.repo'
+fn create_build_script(address string, config BuildConfig, build_arch string) string {
+	repo_url := '$address/$config.repo'
 
 	mut commands := [
 		// This will later be replaced by a proper setting for changing the
 		// mirrorlist
-		"echo -e '[$target.repo]\\nServer = $address/\$repo/\$arch\\nSigLevel = Optional' >> /etc/pacman.conf"
+		"echo -e '[$config.repo]\\nServer = $address/\$repo/\$arch\\nSigLevel = Optional' >> /etc/pacman.conf"
 		// We need to update the package list of the repo we just added above.
 		// This should however not pull in a lot of packages as long as the
 		// builder image is rebuilt frequently.
@@ -38,22 +36,22 @@ fn create_build_script(address string, target &Target, build_arch string) string
 		'su builder',
 	]
 
-	commands << match target.kind {
+	commands << match config.kind {
 		'git' {
-			if target.branch == '' {
+			if config.branch == '' {
 				[
-					"git clone --single-branch --depth 1 '$target.url' repo",
+					"git clone --single-branch --depth 1 '$config.url' repo",
 				]
 			} else {
 				[
-					"git clone --single-branch --depth 1 --branch $target.branch '$target.url' repo",
+					"git clone --single-branch --depth 1 --branch $config.branch '$config.url' repo",
 				]
 			}
 		}
 		'url' {
 			[
 				'mkdir repo',
-				"curl -o repo/PKGBUILD -L '$target.url'",
+				"curl -o repo/PKGBUILD -L '$config.url'",
 			]
 		}
 		else {
@@ -65,14 +63,22 @@ fn create_build_script(address string, target &Target, build_arch string) string
 		'cd repo',
 		'makepkg --nobuild --syncdeps --needed --noconfirm',
 		'source PKGBUILD',
+	]
+
+	if !config.force {
 		// The build container checks whether the package is already present on
 		// the server.
-		'curl -s --head --fail $repo_url/$build_arch/\$pkgname-\$pkgver-\$pkgrel && exit 0',
-		// If the above curl command succeeds, we don't need to rebuild the
-		// package. However, because we're in a su shell, the exit command will
-		// drop us back into the root shell. Therefore, we must check whether
-		// we're in root so we don't proceed.
-		'[ "\$(id -u)" == 0 ] && exit 0',
+		commands << [
+			'curl -s --head --fail $repo_url/$build_arch/\$pkgname-\$pkgver-\$pkgrel && exit 0',
+			// If the above curl command succeeds, we don't need to rebuild the
+			// package. However, because we're in a su shell, the exit command will
+			// drop us back into the root shell. Therefore, we must check whether
+			// we're in root so we don't proceed.
+			'[ "\$(id -u)" == 0 ] && exit 0',
+		]
+	}
+
+	commands << [
 		'MAKEFLAGS="-j\$(nproc)" makepkg -s --noconfirm --needed && for pkg in \$(ls -1 *.pkg*); do curl -XPOST -T "\$pkg" -H "X-API-KEY: \$API_KEY" $repo_url/publish; done',
 	]
 
diff --git a/src/build/shell_test.v b/src/build/shell_test.v
index 341df88..8bb22d9 100644
--- a/src/build/shell_test.v
+++ b/src/build/shell_test.v
@@ -1,42 +1,46 @@
 module build
 
-import models { Target }
-
 fn test_create_build_script_git_branch() {
-	target := Target{
-		id: 1
+	config := BuildConfig{
+		target_id: 1
 		kind: 'git'
 		url: 'https://examplerepo.com'
 		branch: 'main'
 		repo: 'vieter'
+		base_image: 'not-used:latest'
 	}
-	build_script := create_build_script('https://example.com', target, 'x86_64')
+
+	build_script := create_build_script('https://example.com', config, 'x86_64')
 	expected := $embed_file('build_script_git_branch.sh')
 
 	assert build_script == expected.to_string().trim_space()
 }
 
 fn test_create_build_script_git() {
-	target := Target{
-		id: 1
+	config := BuildConfig{
+		target_id: 1
 		kind: 'git'
 		url: 'https://examplerepo.com'
 		repo: 'vieter'
+		base_image: 'not-used:latest'
 	}
-	build_script := create_build_script('https://example.com', target, 'x86_64')
+
+	build_script := create_build_script('https://example.com', config, 'x86_64')
 	expected := $embed_file('build_script_git.sh')
 
 	assert build_script == expected.to_string().trim_space()
 }
 
 fn test_create_build_script_url() {
-	target := Target{
-		id: 1
+	config := BuildConfig{
+		target_id: 1
 		kind: 'url'
 		url: 'https://examplerepo.com'
 		repo: 'vieter'
+		base_image: 'not-used:latest'
 	}
-	build_script := create_build_script('https://example.com', target, 'x86_64')
+
+	build_script := create_build_script('https://example.com', config, 'x86_64')
 	expected := $embed_file('build_script_url.sh')
 
 	assert build_script == expected.to_string().trim_space()
diff --git a/src/client/jobs.v b/src/client/jobs.v
new file mode 100644
index 0000000..440affa
--- /dev/null
+++ b/src/client/jobs.v
@@ -0,0 +1,26 @@
+module client
+
+import build { BuildConfig }
+import web.response { Response }
+
+// poll_jobs requests a list of new build jobs from the server.
+pub fn (c &Client) poll_jobs(arch string, max int) ![]BuildConfig {
+	data := c.send_request<[]BuildConfig>(.get, '/api/v1/jobs/poll', {
+		'arch': arch
+		'max':  max.str()
+	})!
+
+	return data.data
+}
+
+// queue_job adds a new one-time build job for the given target to the job
+// queue.
+pub fn (c &Client) queue_job(target_id int, arch string, force bool) !Response<string> {
+	data := c.send_request<string>(.post, '/api/v1/jobs/queue', {
+		'target': target_id.str()
+		'arch':   arch
+		'force':  force.str()
+	})!
+
+	return data
+}
diff --git a/src/console/targets/build.v b/src/console/targets/build.v
index 9368558..e18077d 100644
--- a/src/console/targets/build.v
+++ b/src/console/targets/build.v
@@ -6,7 +6,7 @@ import os
 import build
 
 // build locally builds the target with the given id.
-fn build(conf Config, target_id int) ! {
+fn build(conf Config, target_id int, force bool) ! {
 	c := client.new(conf.address, conf.api_key)
 	target := c.get_target(target_id)!
 
@@ -16,7 +16,7 @@ fn build(conf Config, target_id int) ! {
 	image_id := build.create_build_image(conf.base_image)!
 
 	println('Running build...')
-	res := build.build_target(conf.address, conf.api_key, image_id, target)!
+	res := build.build_target(conf.address, conf.api_key, image_id, target, force)!
 
 	println('Removing build image...')
 
diff --git a/src/console/targets/targets.v b/src/console/targets/targets.v
index 4179363..b527896 100644
--- a/src/console/targets/targets.v
+++ b/src/console/targets/targets.v
@@ -182,11 +182,44 @@ pub fn cmd() cli.Command {
 				required_args: 1
 				usage: 'id'
 				description: 'Build the target with the given id & publish it.'
+				flags: [
+					cli.Flag{
+						name: 'force'
+						description: 'Build the target without checking whether it needs to be renewed.'
+						flag: cli.FlagType.bool
+					},
+					cli.Flag{
+						name: 'remote'
+						description: 'Schedule the build on the server instead of running it locally.'
+						flag: cli.FlagType.bool
+					},
+					cli.Flag{
+						name: 'arch'
+						description: 'Architecture to schedule build for. Required when using -remote.'
+						flag: cli.FlagType.string
+					},
+				]
 				execute: fn (cmd cli.Command) ! {
 					config_file := cmd.flags.get_string('config-file')!
 					conf := vconf.load<Config>(prefix: 'VIETER_', default_path: config_file)!
 
-					build(conf, cmd.args[0].int())!
+					remote := cmd.flags.get_bool('remote')!
+					force := cmd.flags.get_bool('force')!
+					target_id := cmd.args[0].int()
+
+					if remote {
+						arch := cmd.flags.get_string('arch')!
+
+						if arch == '' {
+							return error('When scheduling the build remotely, you have to specify an architecture.')
+						}
+
+						c := client.new(conf.address, conf.api_key)
+						res := c.queue_job(target_id, arch, force)!
+						println(res.message)
+					} else {
+						build(conf, target_id, force)!
+					}
 				}
 			},
 		]
diff --git a/src/cron/daemon/build.v b/src/cron/daemon/build.v
index beed9fc..42edc92 100644
--- a/src/cron/daemon/build.v
+++ b/src/cron/daemon/build.v
@@ -79,7 +79,7 @@ fn (mut d Daemon) run_build(build_index int, sb ScheduledBuild) {
 	mut status := 0
 
 	res := build.build_target(d.client.address, d.client.api_key, d.builder_images.last(),
-		&sb.target) or {
+		&sb.target, false) or {
 		d.ldebug('build_target error: $err.msg()')
 		status = 1
 
diff --git a/src/db/logs.v b/src/db/logs.v
index 923dde2..2745467 100644
--- a/src/db/logs.v
+++ b/src/db/logs.v
@@ -84,6 +84,8 @@ pub fn (db &VieterDb) add_build_log(log BuildLog) int {
 		insert log into BuildLog
 	}
 
+	// Here, this does work because a log doesn't contain any foreign keys,
+	// meaning the ORM only has to do a single add
 	inserted_id := db.conn.last_id() as int
 
 	return inserted_id
diff --git a/src/db/targets.v b/src/db/targets.v
index a705ebb..41e56df 100644
--- a/src/db/targets.v
+++ b/src/db/targets.v
@@ -38,14 +38,17 @@ pub fn (db &VieterDb) get_target(target_id int) ?Target {
 }
 
 // add_target inserts the given target into the database.
-pub fn (db &VieterDb) add_target(repo Target) int {
+pub fn (db &VieterDb) add_target(target Target) int {
 	sql db.conn {
-		insert repo into Target
+		insert target into Target
 	}
 
-	inserted_id := db.conn.last_id() as int
+	// ID of inserted target is the largest id
+	inserted_target := sql db.conn {
+		select from Target order by id desc limit 1
+	}
 
-	return inserted_id
+	return inserted_target.id
 }
 
 // delete_target deletes the target with the given id from the database.
diff --git a/src/main.v b/src/main.v
index fc09f7e..34387bf 100644
--- a/src/main.v
+++ b/src/main.v
@@ -9,6 +9,7 @@ import console.schedule
 import console.man
 import console.aur
 import cron
+import agent
 
 fn main() {
 	mut app := cli.Command{
@@ -40,6 +41,7 @@ fn main() {
 			schedule.cmd(),
 			man.cmd(),
 			aur.cmd(),
+			agent.cmd(),
 		]
 	}
 	app.setup()
diff --git a/src/server/api_jobs.v b/src/server/api_jobs.v
new file mode 100644
index 0000000..7795351
--- /dev/null
+++ b/src/server/api_jobs.v
@@ -0,0 +1,49 @@
+module server
+
+import web
+import web.response { new_data_response, new_response }
+
+// v1_poll_job_queue allows agents to poll for new build jobs.
+['/api/v1/jobs/poll'; auth; get]
+fn (mut app App) v1_poll_job_queue() web.Result {
+	arch := app.query['arch'] or {
+		return app.json(.bad_request, new_response('Missing arch query arg.'))
+	}
+
+	max_str := app.query['max'] or {
+		return app.json(.bad_request, new_response('Missing max query arg.'))
+	}
+	max := max_str.int()
+
+	mut out := app.job_queue.pop_n(arch, max).map(it.config)
+
+	return app.json(.ok, new_data_response(out))
+}
+
+// v1_queue_job allows queueing a new one-time build job for the given target.
+['/api/v1/jobs/queue'; auth; post]
+fn (mut app App) v1_queue_job() web.Result {
+	target_id := app.query['target'] or {
+		return app.json(.bad_request, new_response('Missing target query arg.'))
+	}.int()
+
+	arch := app.query['arch'] or {
+		return app.json(.bad_request, new_response('Missing arch query arg.'))
+	}
+
+	if arch == '' {
+		app.json(.bad_request, new_response('Empty arch query arg.'))
+	}
+
+	force := 'force' in app.query
+
+	target := app.db.get_target(target_id) or {
+		return app.json(.bad_request, new_response('Unknown target id.'))
+	}
+
+	app.job_queue.insert(target: target, arch: arch, single: true, now: true, force: force) or {
+		return app.status(.internal_server_error)
+	}
+
+	return app.status(.ok)
+}
diff --git a/src/server/api_targets.v b/src/server/api_targets.v
index 16db7e9..dc39d37 100644
--- a/src/server/api_targets.v
+++ b/src/server/api_targets.v
@@ -12,17 +12,17 @@ fn (mut app App) v1_get_targets() web.Result {
 	filter := models.from_params<TargetFilter>(app.query) or {
 		return app.json(http.Status.bad_request, new_response('Invalid query parameters.'))
 	}
-	repos := app.db.get_targets(filter)
+	targets := app.db.get_targets(filter)
 
-	return app.json(.ok, new_data_response(repos))
+	return app.json(.ok, new_data_response(targets))
 }
 
 // v1_get_single_target returns the information for a single target.
 ['/api/v1/targets/:id'; auth; get]
 fn (mut app App) v1_get_single_target(id int) web.Result {
-	repo := app.db.get_target(id) or { return app.not_found() }
+	target := app.db.get_target(id) or { return app.not_found() }
 
-	return app.json(.ok, new_data_response(repo))
+	return app.json(.ok, new_data_response(target))
 }
 
 // v1_post_target creates a new target from the provided query string.
@@ -30,22 +30,27 @@ fn (mut app App) v1_get_single_target(id int) web.Result {
 fn (mut app App) v1_post_target() web.Result {
 	mut params := app.query.clone()
 
-	// If a repo is created without specifying the arch, we assume it's meant
+	// If a target is created without specifying the arch, we assume it's meant
 	// for the default architecture.
 	if 'arch' !in params || params['arch'] == '' {
 		params['arch'] = app.conf.default_arch
 	}
 
-	new_repo := models.from_params<Target>(params) or {
+	mut new_target := models.from_params<Target>(params) or {
 		return app.json(http.Status.bad_request, new_response(err.msg()))
 	}
 
 	// Ensure someone doesn't submit an invalid kind
-	if new_repo.kind !in models.valid_kinds {
+	if new_target.kind !in models.valid_kinds {
 		return app.json(http.Status.bad_request, new_response('Invalid kind.'))
 	}
 
-	id := app.db.add_target(new_repo)
+	id := app.db.add_target(new_target)
+	new_target.id = id
+
+	// Add the target to the job queue
+	// TODO return better error here if it's the cron schedule that's incorrect
+	app.job_queue.insert_all(new_target) or { return app.status(.internal_server_error) }
 
 	return app.json(.ok, new_data_response(id))
 }
@@ -54,6 +59,7 @@ fn (mut app App) v1_post_target() web.Result {
 ['/api/v1/targets/:id'; auth; delete]
 fn (mut app App) v1_delete_target(id int) web.Result {
 	app.db.delete_target(id)
+	app.job_queue.invalidate(id)
 
 	return app.json(.ok, new_response(''))
 }
@@ -69,7 +75,10 @@ fn (mut app App) v1_patch_target(id int) web.Result {
 		app.db.update_target_archs(id, arch_objs)
 	}
 
-	repo := app.db.get_target(id) or { return app.status(.internal_server_error) }
+	target := app.db.get_target(id) or { return app.status(.internal_server_error) }
 
-	return app.json(.ok, new_data_response(repo))
+	app.job_queue.invalidate(id)
+	app.job_queue.insert_all(target) or { return app.status(.internal_server_error) }
+
+	return app.json(.ok, new_data_response(target))
 }
diff --git a/src/server/cli.v b/src/server/cli.v
index a9644f3..2fede6c 100644
--- a/src/server/cli.v
+++ b/src/server/cli.v
@@ -5,12 +5,14 @@ import conf as vconf
 
 struct Config {
 pub:
-	log_level    string = 'WARN'
-	pkg_dir      string
-	data_dir     string
-	api_key      string
-	default_arch string
-	port         int = 8000
+	log_level       string = 'WARN'
+	pkg_dir         string
+	data_dir        string
+	api_key         string
+	default_arch    string
+	global_schedule string = '0 3'
+	port            int    = 8000
+	base_image      string = 'archlinux:base-devel'
 }
 
 // cmd returns the cli submodule that handles starting the server
diff --git a/src/server/server.v b/src/server/server.v
index d5f6135..74b1f37 100644
--- a/src/server/server.v
+++ b/src/server/server.v
@@ -6,6 +6,8 @@ import log
 import repo
 import util
 import db
+import build { BuildJobQueue }
+import cron.expression
 
 const (
 	log_file_name = 'vieter.log'
@@ -20,7 +22,26 @@ pub:
 	conf Config [required; web_global]
 pub mut:
 	repo repo.RepoGroupManager [required; web_global]
-	db   db.VieterDb
+	// Keys are the various architectures for packages
+	job_queue BuildJobQueue [required; web_global]
+	db        db.VieterDb
+}
+
+// init_job_queue populates a fresh job queue with all the targets currently
+// stored in the database.
+fn (mut app App) init_job_queue() ! {
+	// Initialize build queues
+	mut targets := app.db.get_targets(limit: 25)
+	mut i := u64(0)
+
+	for targets.len > 0 {
+		for target in targets {
+			app.job_queue.insert_all(target)!
+		}
+
+		i += 25
+		targets = app.db.get_targets(limit: 25, offset: i)
+	}
 }
 
 // server starts the web server & starts listening for requests
@@ -30,6 +51,10 @@ pub fn server(conf Config) ! {
 		util.exit_with_message(1, "'any' is not allowed as the value for default_arch.")
 	}
 
+	global_ce := expression.parse_expression(conf.global_schedule) or {
+		util.exit_with_message(1, 'Invalid global cron expression: $err.msg()')
+	}
+
 	// Configure logger
 	log_level := log.level_from_tag(conf.log_level) or {
 		util.exit_with_message(1, 'Invalid log level. The allowed values are FATAL, ERROR, WARN, INFO & DEBUG.')
@@ -71,11 +96,17 @@ pub fn server(conf Config) ! {
 		util.exit_with_message(1, 'Failed to initialize database: $err.msg()')
 	}
 
-	web.run(&App{
+	mut app := &App{
 		logger: logger
 		api_key: conf.api_key
 		conf: conf
 		repo: repo
 		db: db
-	}, conf.port)
+		job_queue: build.new_job_queue(global_ce, conf.base_image)
+	}
+	app.init_job_queue() or {
+		util.exit_with_message(1, 'Failed to inialize job queue: $err.msg()')
+	}
+
+	web.run(app, conf.port)
 }
diff --git a/vieter.toml b/vieter.toml
index d3922a4..74a7397 100644
--- a/vieter.toml
+++ b/vieter.toml
@@ -4,11 +4,11 @@ data_dir = "data"
 pkg_dir = "data/pkgs"
 log_level = "DEBUG"
 default_arch = "x86_64"
+arch = "x86_64"
 
 address = "http://localhost:8000"
 
-global_schedule = '* *'
+# global_schedule = '* *'
 api_update_frequency = 2
 image_rebuild_frequency = 1
 max_concurrent_builds = 3
-