diff --git a/include/sched.h b/include/sched.h
new file mode 100644
index 0000000..e421e4b
--- /dev/null
+++ b/include/sched.h
@@ -0,0 +1,45 @@
+#ifndef FROSTBYTE_SCHED_H
+#define FROSTBYTE_SCHED_H
+
+#include "types.h"
+#include "vfs.h"
+
+typedef enum {
+    threadUnused = 0,
+    threadRunnable = 1,
+    threadRunning = 2,
+    threadSleeping = 3,
+    threadDead = 4
+} ThreadState;
+
+typedef void (*ThreadEntry)(void);
+
+typedef struct Thread {
+    uint64 rsp;
+    uint64 stackBase;
+    uint64 kernelStackTop;
+    uint64 sleepUntil;
+    uint64 brk;
+    uint32 id;
+    ThreadState state;
+    const char* name;
+    FileDescriptor fds[maxOpenFiles];
+} Thread;
+
+void schedInit(void);
+uint32 threadCreate(const char* name, ThreadEntry entry);
+uint32 threadCreateUser(const char* name, uint64 entry, uint64 userStackTop);
+uint64 schedule(uint64 currentRsp);
+void schedStart(void);
+
+void threadExit(uint64 code);
+void threadSleep(uint64 milliseconds);
+void threadSetBrk(uint32 id, uint64 brk);
+uint64 currentThreadSbrk(int64 increment);
+
+uint32 threadCount(void);
+Thread* threadAt(uint32 index);
+Thread* currentThread(void);
+uint32 currentThreadId(void);
+
+#endif
diff --git a/src/sched.c b/src/sched.c
new file mode 100644
index 0000000..ad6c639
--- /dev/null
+++ b/src/sched.c
@@ -0,0 +1,183 @@
+#include "sched.h"
+#include "isr.h"
+#include "heap.h"
+#include "string.h"
+#include "ports.h"
+#include "gdt.h"
+#include "tss.h"
+#include "pit.h"
+#include "console.h"
+#include "paging.h"
+#include "vfs.h"
+
+#define maxThreads 16
+#define threadStackSize 16384
+
+static Thread threads[maxThreads];
+static uint32 threadTotal;
+static uint32 currentIndex;
+static bool schedulerActive;
+
+void schedInit(void)
+{
+    threadTotal = 1;
+    currentIndex = 0;
+
+    threads[0].id = 0;
+    threads[0].name = "idle";
+    threads[0].state = threadRunning;
+    threads[0].rsp = 0;
+    threads[0].stackBase = 0;
+    threads[0].kernelStackTop = 0;
+    threads[0].sleepUntil = 0;
+    threads[0].brk = 0;
+    vfsFdInit(threads[0].fds);
+}
+
+uint32 threadCreate(const char* name, ThreadEntry entry)
+{
+    uint32 slot = threadTotal;
+    threadTotal++;
+
+    uint64 stackBase = (uint64)kmalloc(threadStackSize);
+    uint64 stackTop = (stackBase + threadStackSize) & ~15ull;
+
+    Registers* frame = (Registers*)(stackTop - sizeof(Registers));
+    memset(frame, 0, sizeof(Registers));
+    frame->rip = (uint64)entry;
+    frame->cs = kernelCodeSelector;
+    frame->rflags = 0x202;
+    frame->rsp = stackTop;
+    frame->ss = kernelDataSelector;
+
+    threads[slot].id = slot;
+    threads[slot].name = name;
+    threads[slot].state = threadRunnable;
+    threads[slot].stackBase = stackBase;
+    threads[slot].kernelStackTop = stackTop;
+    threads[slot].rsp = (uint64)frame;
+    threads[slot].sleepUntil = 0;
+    threads[slot].brk = 0;
+    vfsFdInit(threads[slot].fds);
+
+    return slot;
+}
+
+uint32 threadCreateUser(const char* name, uint64 entry, uint64 userStackTop)
+{
+    uint32 slot = threadTotal;
+    threadTotal++;
+
+    uint64 stackBase = (uint64)kmalloc(threadStackSize);
+    uint64 kernelStackTop = (stackBase + threadStackSize) & ~15ull;
+
+    Registers* frame = (Registers*)(kernelStackTop - sizeof(Registers));
+    memset(frame, 0, sizeof(Registers));
+    frame->rip = entry;
+    frame->cs = userCodeSelector | 3;
+    frame->rflags = 0x202;
+    frame->rsp = userStackTop;
+    frame->ss = userDataSelector | 3;
+
+    threads[slot].id = slot;
+    threads[slot].name = name;
+    threads[slot].state = threadRunnable;
+    threads[slot].stackBase = stackBase;
+    threads[slot].kernelStackTop = kernelStackTop;
+    threads[slot].rsp = (uint64)frame;
+    threads[slot].sleepUntil = 0;
+    threads[slot].brk = 0;
+    vfsFdInit(threads[slot].fds);
+
+    return slot;
+}
+
+uint64 schedule(uint64 currentRsp)
+{
+    if (!schedulerActive)
+        return currentRsp;
+
+    threads[currentIndex].rsp = currentRsp;
+    if (threads[currentIndex].state == threadRunning)
+        threads[currentIndex].state = threadRunnable;
+
+    uint64 now = pitTicks();
+    for (uint32 i = 0; i < threadTotal; i++)
+        if (threads[i].state == threadSleeping && now >= threads[i].sleepUntil)
+            threads[i].state = threadRunnable;
+
+    uint32 next = 0;
+    for (uint32 step = 1; step <= threadTotal; step++) {
+        uint32 candidate = (currentIndex + step) % threadTotal;
+        if (threads[candidate].state == threadRunnable) {
+            next = candidate;
+            break;
+        }
+    }
+
+    currentIndex = next;
+    threads[next].state = threadRunning;
+    tssSetRsp0(threads[next].kernelStackTop);
+    return threads[next].rsp;
+}
+
+void schedStart(void)
+{
+    schedulerActive = true;
+    enableInterrupts();
+    for (;;)
+        halt();
+}
+
+void threadExit(uint64 code)
+{
+    Thread* thread = &threads[currentIndex];
+    thread->state = threadDead;
+    kprintf("[user] pid=%u '%s' exited with code %lu\n", thread->id, thread->name, code);
+}
+
+void threadSleep(uint64 milliseconds)
+{
+    uint64 ticks = milliseconds * pitFrequency() / 1000;
+    if (ticks == 0)
+        ticks = 1;
+    threads[currentIndex].sleepUntil = pitTicks() + ticks;
+    threads[currentIndex].state = threadSleeping;
+}
+
+void threadSetBrk(uint32 id, uint64 brk)
+{
+    threads[id].brk = brk;
+}
+
+uint64 currentThreadSbrk(int64 increment)
+{
+    Thread* thread = &threads[currentIndex];
+    uint64 previous = thread->brk;
+
+    if (increment > 0)
+        pagingMapUserRange(previous, (uint64)increment);
+
+    thread->brk = previous + (uint64)increment;
+    return previous;
+}
+
+uint32 threadCount(void)
+{
+    return threadTotal;
+}
+
+Thread* threadAt(uint32 index)
+{
+    return &threads[index];
+}
+
+Thread* currentThread(void)
+{
+    return &threads[currentIndex];
+}
+
+uint32 currentThreadId(void)
+{
+    return threads[currentIndex].id;
+}