diff --git a/.woodpecker/.test.yml b/.woodpecker/.test.yml
new file mode 100644
index 0000000..ec559c8
--- /dev/null
+++ b/.woodpecker/.test.yml
@@ -0,0 +1,15 @@
+matrix:
+  PLATFORM:
+    - linux/amd64
+    - linux/arm64
+
+platform: ${PLATFORM}
+
+pipeline:
+  test:
+    image: 'chewingbever/vlang:latest'
+    pull: true
+    commands:
+      - make test
+    when:
+      event: push
diff --git a/Makefile b/Makefile
index 76ab7b5..9421fb6 100644
--- a/Makefile
+++ b/Makefile
@@ -60,6 +60,10 @@ fmt:
 vet:
 	$(V) vet -W $(SRC_DIR)
 
+.PHONY: test
+test:
+	$(V) test $(SRC_DIR)
+
 # Build & patch the V compiler
 .PHONY: v
 v: v/v
diff --git a/src/cron/cli.v b/src/cron/cli.v
new file mode 100644
index 0000000..8e6b0f1
--- /dev/null
+++ b/src/cron/cli.v
@@ -0,0 +1,27 @@
+module cron
+
+import cli
+import env
+
+struct Config {
+pub:
+	log_level  string = 'WARN'
+	log_file   string = 'vieter.log'
+	api_key    string
+	address    string
+	base_image string = 'archlinux:base-devel'
+}
+
+// cmd returns the cli module that handles the cron daemon.
+pub fn cmd() cli.Command {
+	return cli.Command{
+		name: 'cron'
+		description: 'Start the cron service that periodically runs builds.'
+		execute: fn (cmd cli.Command) ? {
+			config_file := cmd.flags.get_string('config-file') ?
+			conf := env.load<Config>(config_file) ?
+
+			cron(conf) ?
+		}
+	}
+}
diff --git a/src/cron/cron.v b/src/cron/cron.v
new file mode 100644
index 0000000..3ba9d0f
--- /dev/null
+++ b/src/cron/cron.v
@@ -0,0 +1,18 @@
+module cron
+
+import git
+import time
+
+struct ScheduledBuild {
+	repo      git.GitRepo
+	timestamp time.Time
+}
+
+fn (r1 ScheduledBuild) < (r2 ScheduledBuild) bool {
+	return r1.timestamp < r2.timestamp
+}
+
+// cron starts a cron daemon & starts periodically scheduling builds.
+pub fn cron(conf Config) ? {
+	println('WIP')
+}
diff --git a/src/cron/expression.v b/src/cron/expression.v
new file mode 100644
index 0000000..0a35541
--- /dev/null
+++ b/src/cron/expression.v
@@ -0,0 +1,254 @@
+module cron
+
+import time
+
+struct CronExpression {
+	minutes []int
+	hours   []int
+	days    []int
+	months  []int
+}
+
+// next calculates the earliest time this cron expression is valid. It will
+// always pick a moment in the future, even if ref matches completely up to the
+// minute. This function conciously does not take gap years into account.
+pub fn (ce &CronExpression) next(ref time.Time) ?time.Time {
+	// If the given ref matches the next cron occurence up to the minute, it
+	// will return that value. Because we always want to return a value in the
+	// future, we artifically shift the ref 60 seconds to make sure we always
+	// match in the future. A shift of 60 seconds is enough because the cron
+	// expression does not allow for accuracy smaller than one minute.
+	sref := ref
+
+	// For all of these values, the rule is the following: if their value is
+	// the length of their respective array in the CronExpression object, that
+	// means we've looped back around. This means that the "bigger" value has
+	// to be incremented by one. For example, if the minutes have looped
+	// around, that means that the hour has to be incremented as well.
+	mut minute_index := 0
+	mut hour_index := 0
+	mut day_index := 0
+	mut month_index := 0
+
+	// This chain is the same logic multiple times, namely that if a "bigger"
+	// value loops around, then the smaller value will always reset as well.
+	// For example, if we're going to a new day, the hour & minute will always
+	// be their smallest value again.
+	for month_index < ce.months.len && sref.month > ce.months[month_index] {
+		month_index++
+	}
+
+	if month_index < ce.months.len && sref.month == ce.months[month_index] {
+		for day_index < ce.days.len && sref.day > ce.days[day_index] {
+			day_index++
+		}
+
+		if day_index < ce.days.len && ce.days[day_index] == sref.day {
+			for hour_index < ce.hours.len && sref.hour > ce.hours[hour_index] {
+				hour_index++
+			}
+
+			if hour_index < ce.hours.len && ce.hours[hour_index] == sref.hour {
+				// Minute is the only value where we explicitely make sure we
+				// can't match sref's value exactly. This is to ensure we only
+				// return values in the future.
+				for minute_index < ce.minutes.len && sref.minute >= ce.minutes[minute_index] {
+					minute_index++
+				}
+			}
+		}
+	}
+
+	// Here, we increment the "bigger" values by one if the smaller ones loop
+	// around. The order is important, as it allows a sort-of waterfall effect
+	// to occur which updates all values if required.
+	if minute_index == ce.minutes.len && hour_index < ce.hours.len {
+		hour_index += 1
+	}
+	if hour_index == ce.hours.len && day_index < ce.days.len {
+		day_index += 1
+	}
+
+	if day_index == ce.days.len && month_index < ce.months.len {
+		month_index += 1
+	}
+
+	mut minute := ce.minutes[minute_index % ce.minutes.len]
+	mut hour := ce.hours[hour_index % ce.hours.len]
+	mut day := ce.days[day_index % ce.days.len]
+
+	// Sometimes, we end up with a day that does not exist within the selected
+	// month, e.g. day 30 in February. When this occurs, we reset day back to
+	// the smallest value & loop over to the next month that does have this
+	// day.
+	if day > time.month_days[ce.months[month_index % ce.months.len] - 1] {
+		day = ce.days[0]
+		month_index += 1
+
+		for day > time.month_days[ce.months[month_index & ce.months.len] - 1] {
+			month_index += 1
+
+			// If for whatever reason the day value ends up being something
+			// that can't be scheduled in any month, we have to make sure we
+			// don't create an infinite loop.
+			if month_index == 2 * ce.months.len {
+				return error('No schedulable moment.')
+			}
+		}
+	}
+
+	month := ce.months[month_index % ce.months.len]
+	mut year := sref.year
+
+	// If the month loops over, we need to increment the year.
+	if month_index >= ce.months.len {
+		year++
+	}
+
+	return time.new_time(time.Time{
+		year: year
+		month: month
+		day: day
+		minute: minute
+		hour: hour
+	})
+}
+
+fn (ce &CronExpression) next_from_now() ?time.Time {
+	return ce.next(time.now())
+}
+
+// parse_range parses a given string into a range of sorted integers, if
+// possible.
+fn parse_range(s string, min int, max int, mut bitv []bool) ? {
+	mut start := min
+	mut end := max
+	mut interval := 1
+
+	exps := s.split('/')
+
+	if exps.len > 2 {
+		return error('Invalid expression.')
+	}
+
+	if exps[0] != '*' {
+		dash_parts := exps[0].split('-')
+
+		if dash_parts.len > 2 {
+			return error('Invalid expression.')
+		}
+
+		start = dash_parts[0].int()
+
+		// The builtin parsing functions return zero if the string can't be
+		// parsed into a number, so we have to explicitely check whether they
+		// actually entered zero or if it's an invalid number.
+		if start == 0 && dash_parts[0] != '0' {
+			return error('Invalid number.')
+		}
+
+		// Check whether the start value is out of range
+		if start < min || start > max {
+			return error('Out of range.')
+		}
+
+		if dash_parts.len == 2 {
+			end = dash_parts[1].int()
+
+			if end == 0 && dash_parts[1] != '0' {
+				return error('Invalid number.')
+			}
+
+			if end < start || end > max {
+				return error('Out of range.')
+			}
+		}
+	}
+
+	if exps.len > 1 {
+		interval = exps[1].int()
+
+		// interval being zero is always invalid, but we want to check why
+		// it's invalid for better error messages.
+		if interval == 0 {
+			if exps[1] != '0' {
+				return error('Invalid number.')
+			} else {
+				return error('Step size zero not allowed.')
+			}
+		}
+
+		if interval > max - min {
+			return error('Step size too large.')
+		}
+	}
+	// Here, s solely consists of a number, so that's the only value we
+	// should return.
+	else if exps[0] != '*' && !exps[0].contains('-') {
+		bitv[start - min] = true
+		return
+	}
+
+	for start <= end {
+		bitv[start - min] = true
+		start += interval
+	}
+}
+
+fn bitv_to_ints(bitv []bool, min int) []int {
+	mut out := []int{}
+
+	for i in 0 .. bitv.len {
+		if bitv[i] {
+			out << min + i
+		}
+	}
+
+	return out
+}
+
+fn parse_part(s string, min int, max int) ?[]int {
+	mut bitv := []bool{len: max - min + 1, init: false}
+
+	for range in s.split(',') {
+		parse_range(range, min, max, mut bitv) ?
+	}
+
+	return bitv_to_ints(bitv, min)
+}
+
+// parse_expression parses an entire cron expression string into a
+// CronExpression object, if possible.
+fn parse_expression(exp string) ?CronExpression {
+	// The filter allows for multiple spaces between parts
+	mut parts := exp.split(' ').filter(it != '')
+
+	if parts.len < 2 || parts.len > 4 {
+		return error('Expression must contain between 2 and 4 space-separated parts.')
+	}
+
+	// For ease of use, we allow the user to only specify as many parts as they
+	// need.
+	for parts.len < 4 {
+		parts << '*'
+	}
+
+	mut part_results := [][]int{}
+
+	mins := [0, 0, 1, 1]
+	maxs := [59, 23, 31, 12]
+
+	// This for loop allows us to more clearly propagate the error to the user.
+	for i, min in mins {
+		part_results << parse_part(parts[i], min, maxs[i]) or {
+			return error('An error occurred with part $i: $err.msg')
+		}
+	}
+
+	return CronExpression{
+		minutes: part_results[0]
+		hours: part_results[1]
+		days: part_results[2]
+		months: part_results[3]
+	}
+}
diff --git a/src/cron/expression_parse_test.v b/src/cron/expression_parse_test.v
new file mode 100644
index 0000000..8f3ac38
--- /dev/null
+++ b/src/cron/expression_parse_test.v
@@ -0,0 +1,98 @@
+module cron
+
+// parse_range_error returns the returned error message. If the result is '',
+// that means the function didn't error.
+fn parse_range_error(s string, min int, max int) string {
+	mut bitv := []bool{len: max - min + 1, init: false}
+
+	parse_range(s, min, max, mut bitv) or { return err.msg }
+
+	return ''
+}
+
+// =====parse_range=====
+fn test_range_star_range() ? {
+	mut bitv := []bool{len: 6, init: false}
+	parse_range('*', 0, 5, mut bitv) ?
+
+	assert bitv == [true, true, true, true, true, true]
+}
+
+fn test_range_number() ? {
+	mut bitv := []bool{len: 6, init: false}
+	parse_range('4', 0, 5, mut bitv) ?
+
+	assert bitv_to_ints(bitv, 0) == [4]
+}
+
+fn test_range_number_too_large() ? {
+	assert parse_range_error('10', 0, 6) == 'Out of range.'
+}
+
+fn test_range_number_too_small() ? {
+	assert parse_range_error('0', 2, 6) == 'Out of range.'
+}
+
+fn test_range_number_invalid() ? {
+	assert parse_range_error('x', 0, 6) == 'Invalid number.'
+}
+
+fn test_range_step_star_1() ? {
+	mut bitv := []bool{len: 21, init: false}
+	parse_range('*/4', 0, 20, mut bitv) ?
+
+	assert bitv_to_ints(bitv, 0) == [0, 4, 8, 12, 16, 20]
+}
+
+fn test_range_step_star_2() ? {
+	mut bitv := []bool{len: 8, init: false}
+	parse_range('*/3', 1, 8, mut bitv) ?
+
+	assert bitv_to_ints(bitv, 1) == [1, 4, 7]
+}
+
+fn test_range_step_star_too_large() ? {
+	assert parse_range_error('*/21', 0, 20) == 'Step size too large.'
+}
+
+fn test_range_step_zero() ? {
+	assert parse_range_error('*/0', 0, 20) == 'Step size zero not allowed.'
+}
+
+fn test_range_step_number() ? {
+	mut bitv := []bool{len: 21, init: false}
+	parse_range('5/4', 2, 22, mut bitv) ?
+
+	assert bitv_to_ints(bitv, 2) == [5, 9, 13, 17, 21]
+}
+
+fn test_range_step_number_too_large() ? {
+	assert parse_range_error('10/4', 0, 5) == 'Out of range.'
+}
+
+fn test_range_step_number_too_small() ? {
+	assert parse_range_error('2/4', 5, 10) == 'Out of range.'
+}
+
+fn test_range_dash() ? {
+	mut bitv := []bool{len: 10, init: false}
+	parse_range('4-8', 0, 9, mut bitv) ?
+
+	assert bitv_to_ints(bitv, 0) == [4, 5, 6, 7, 8]
+}
+
+fn test_range_dash_step() ? {
+	mut bitv := []bool{len: 10, init: false}
+	parse_range('4-8/2', 0, 9, mut bitv) ?
+
+	assert bitv_to_ints(bitv, 0) == [4, 6, 8]
+}
+
+// =====parse_part=====
+fn test_part_single() ? {
+	assert parse_part('*', 0, 5) ? == [0, 1, 2, 3, 4, 5]
+}
+
+fn test_part_multiple() ? {
+	assert parse_part('*/2,2/3', 1, 8) ? == [1, 2, 3, 5, 7, 8]
+}
diff --git a/src/cron/expression_test.v b/src/cron/expression_test.v
new file mode 100644
index 0000000..0be9a64
--- /dev/null
+++ b/src/cron/expression_test.v
@@ -0,0 +1,34 @@
+module cron
+
+import time { parse }
+
+fn util_test_time(exp string, t1_str string, t2_str string) ? {
+	ce := parse_expression(exp) ?
+	t1 := parse(t1_str) ?
+	t2 := parse(t2_str) ?
+
+	t3 := ce.next(t1) ?
+
+	assert t2.year == t3.year
+	assert t2.month == t3.month
+	assert t2.day == t3.day
+	assert t2.hour == t3.hour
+	assert t2.minute == t3.minute
+}
+
+fn test_next_simple() ? {
+	// Very simple
+	util_test_time('0 3', '2002-01-01 00:00:00', '2002-01-01 03:00:00') ?
+
+	// Overlap to next day
+	util_test_time('0 3', '2002-01-01 03:00:00', '2002-01-02 03:00:00') ?
+	util_test_time('0 3', '2002-01-01 04:00:00', '2002-01-02 03:00:00') ?
+
+	util_test_time('0 3/4', '2002-01-01 04:00:00', '2002-01-01 07:00:00') ?
+
+	// Overlap to next month
+	util_test_time('0 3', '2002-11-31 04:00:00', '2002-12-01 03:00:00') ?
+
+	// Overlap to next year
+	util_test_time('0 3', '2002-12-31 04:00:00', '2003-01-01 03:00:00') ?
+}
diff --git a/src/env.v b/src/env/env.v
similarity index 100%
rename from src/env.v
rename to src/env/env.v
diff --git a/src/main.v b/src/main.v
index 3025389..37cabc3 100644
--- a/src/main.v
+++ b/src/main.v
@@ -5,6 +5,7 @@ import server
 import cli
 import build
 import git
+import cron
 
 fn main() {
 	mut app := cli.Command{
@@ -25,6 +26,7 @@ fn main() {
 			server.cmd(),
 			build.cmd(),
 			git.cmd(),
+			cron.cmd(),
 		]
 	}
 
diff --git a/src/package.v b/src/package/package.v
similarity index 100%
rename from src/package.v
rename to src/package/package.v
diff --git a/src/response.v b/src/response/response.v
similarity index 100%
rename from src/response.v
rename to src/response/response.v
diff --git a/src/util.v b/src/util/util.v
similarity index 99%
rename from src/util.v
rename to src/util/util.v
index 49c9d22..228f584 100644
--- a/src/util.v
+++ b/src/util/util.v
@@ -44,6 +44,7 @@ pub fn reader_to_file(mut reader io.BufferedReader, length int, path string) ? {
 		for to_write > 0 {
 			// TODO don't just loop infinitely here
 			bytes_written := file.write(buf[bytes_read - to_write..bytes_read]) or { continue }
+			// file.flush()
 
 			to_write = to_write - bytes_written
 		}