diff --git a/src-self-hosted/Module.zig b/src-self-hosted/Module.zig
index df7c1986f3..5e22058acf 100644
--- a/src-self-hosted/Module.zig
+++ b/src-self-hosted/Module.zig
@@ -1571,7 +1571,7 @@ fn analyzeRootSrcFile(self: *Module, root_scope: *Scope.File) !void {
                         .macho => {
                             // TODO Implement for MachO
                         },
-                        .c => {},
+                        .c, .wasm => {},
                     }
                 }
             } else {
@@ -1781,11 +1781,13 @@ fn allocateNewDecl(
             .elf => .{ .elf = link.File.Elf.TextBlock.empty },
             .macho => .{ .macho = link.File.MachO.TextBlock.empty },
             .c => .{ .c = {} },
+            .wasm => .{ .wasm = {} },
         },
         .fn_link = switch (self.bin_file.tag) {
             .elf => .{ .elf = link.File.Elf.SrcFn.empty },
             .macho => .{ .macho = link.File.MachO.SrcFn.empty },
             .c => .{ .c = {} },
+            .wasm => .{ .wasm = null },
         },
         .generation = 0,
     };
diff --git a/src-self-hosted/codegen/wasm.zig b/src-self-hosted/codegen/wasm.zig
new file mode 100644
index 0000000000..78d8d22ded
--- /dev/null
+++ b/src-self-hosted/codegen/wasm.zig
@@ -0,0 +1,70 @@
+const std = @import("std");
+const Allocator = std.mem.Allocator;
+const ArrayList = std.ArrayList;
+const assert = std.debug.assert;
+const leb = std.debug.leb;
+
+const Decl = @import("../Module.zig").Decl;
+const Type = @import("../type.zig").Type;
+
+fn genValtype(ty: Type) u8 {
+    return switch (ty.tag()) {
+        .u32, .i32 => 0x7F,
+        .u64, .i64 => 0x7E,
+        .f32 => 0x7D,
+        .f64 => 0x7C,
+        else => @panic("TODO: Implement more types for wasm."),
+    };
+}
+
+pub fn genFunctype(buf: *ArrayList(u8), decl: *Decl) !void {
+    const ty = decl.typed_value.most_recent.typed_value.ty;
+    const writer = buf.writer();
+
+    // functype magic
+    try writer.writeByte(0x60);
+
+    // param types
+    try leb.writeULEB128(writer, @intCast(u32, ty.fnParamLen()));
+    if (ty.fnParamLen() != 0) {
+        const params = try buf.allocator.alloc(Type, ty.fnParamLen());
+        defer buf.allocator.free(params);
+        ty.fnParamTypes(params);
+        for (params) |param_type| try writer.writeByte(genValtype(param_type));
+    }
+
+    // return type
+    const return_type = ty.fnReturnType();
+    switch (return_type.tag()) {
+        .void, .noreturn => try leb.writeULEB128(writer, @as(u32, 0)),
+        else => {
+            try leb.writeULEB128(writer, @as(u32, 1));
+            try writer.writeByte(genValtype(return_type));
+        },
+    }
+}
+
+pub fn genCode(buf: *ArrayList(u8), decl: *Decl) !void {
+    assert(buf.items.len == 0);
+    const writer = buf.writer();
+
+    // Reserve space to write the size after generating the code
+    try writer.writeAll(&([1]u8{undefined} ** 5));
+
+    // Write the size of the locals vec
+    // TODO: implement locals
+    try leb.writeULEB128(writer, @as(u32, 0));
+
+    // Write instructions
+
+    // TODO: actually implement codegen
+    try writer.writeByte(0x41); // i32.const
+    try leb.writeILEB128(writer, @as(i32, 42));
+
+    // Write 'end' opcode
+    try writer.writeByte(0x0B);
+
+    // Fill in the size of the generated code to the reserved space at the
+    // beginning of the buffer.
+    leb.writeUnsignedFixed(5, buf.items[0..5], @intCast(u32, buf.items.len - 5));
+}
diff --git a/src-self-hosted/link.zig b/src-self-hosted/link.zig
index a835cc6b7c..6a51138785 100644
--- a/src-self-hosted/link.zig
+++ b/src-self-hosted/link.zig
@@ -46,12 +46,14 @@ pub const File = struct {
         elf: Elf.TextBlock,
         macho: MachO.TextBlock,
         c: void,
+        wasm: void,
     };
 
     pub const LinkFn = union {
         elf: Elf.SrcFn,
         macho: MachO.SrcFn,
         c: void,
+        wasm: ?Wasm.FnData,
     };
 
     tag: Tag,
@@ -69,7 +71,7 @@ pub const File = struct {
             .coff => return error.TODOImplementCoff,
             .elf => return Elf.openPath(allocator, dir, sub_path, options),
             .macho => return MachO.openPath(allocator, dir, sub_path, options),
-            .wasm => return error.TODOImplementWasm,
+            .wasm => return Wasm.openPath(allocator, dir, sub_path, options),
             .c => return C.openPath(allocator, dir, sub_path, options),
             .hex => return error.TODOImplementHex,
             .raw => return error.TODOImplementRaw,
@@ -93,7 +95,7 @@ pub const File = struct {
                     .mode = determineMode(base.options),
                 });
             },
-            .c => {},
+            .c, .wasm => {},
         }
     }
 
@@ -102,6 +104,7 @@ pub const File = struct {
         if (base.file) |f| {
             f.close();
             base.file = null;
+
         }
     }
 
@@ -110,6 +113,7 @@ pub const File = struct {
             .elf => return @fieldParentPtr(Elf, "base", base).updateDecl(module, decl),
             .macho => return @fieldParentPtr(MachO, "base", base).updateDecl(module, decl),
             .c => return @fieldParentPtr(C, "base", base).updateDecl(module, decl),
+            .wasm => return @fieldParentPtr(Wasm, "base", base).updateDecl(module, decl),
         }
     }
 
@@ -117,7 +121,7 @@ pub const File = struct {
         switch (base.tag) {
             .elf => return @fieldParentPtr(Elf, "base", base).updateDeclLineNumber(module, decl),
             .macho => return @fieldParentPtr(MachO, "base", base).updateDeclLineNumber(module, decl),
-            .c => {},
+            .c, .wasm => {},
         }
     }
 
@@ -125,7 +129,7 @@ pub const File = struct {
         switch (base.tag) {
             .elf => return @fieldParentPtr(Elf, "base", base).allocateDeclIndexes(decl),
             .macho => return @fieldParentPtr(MachO, "base", base).allocateDeclIndexes(decl),
-            .c => {},
+            .c, .wasm => {},
         }
     }
 
@@ -135,6 +139,7 @@ pub const File = struct {
             .elf => @fieldParentPtr(Elf, "base", base).deinit(),
             .macho => @fieldParentPtr(MachO, "base", base).deinit(),
             .c => @fieldParentPtr(C, "base", base).deinit(),
+            .wasm => @fieldParentPtr(Wasm, "base", base).deinit(),
         }
     }
 
@@ -155,6 +160,11 @@ pub const File = struct {
                 parent.deinit();
                 base.allocator.destroy(parent);
             },
+            .wasm => {
+                const parent = @fieldParentPtr(Wasm, "base", base);
+                parent.deinit();
+                base.allocator.destroy(parent);
+            },
         }
     }
 
@@ -167,6 +177,7 @@ pub const File = struct {
             .elf => @fieldParentPtr(Elf, "base", base).flush(),
             .macho => @fieldParentPtr(MachO, "base", base).flush(),
             .c => @fieldParentPtr(C, "base", base).flush(),
+            .wasm => @fieldParentPtr(Wasm, "base", base).flush(),
         };
     }
 
@@ -175,6 +186,7 @@ pub const File = struct {
             .elf => @fieldParentPtr(Elf, "base", base).freeDecl(decl),
             .macho => @fieldParentPtr(MachO, "base", base).freeDecl(decl),
             .c => unreachable,
+            .wasm => @fieldParentPtr(Wasm, "base", base).freeDecl(decl),
         }
     }
 
@@ -183,6 +195,7 @@ pub const File = struct {
             .elf => @fieldParentPtr(Elf, "base", base).error_flags,
             .macho => @fieldParentPtr(MachO, "base", base).error_flags,
             .c => return .{ .no_entry_point_found = false },
+            .wasm => return ErrorFlags{},
         };
     }
 
@@ -197,6 +210,7 @@ pub const File = struct {
             .elf => return @fieldParentPtr(Elf, "base", base).updateDeclExports(module, decl, exports),
             .macho => return @fieldParentPtr(MachO, "base", base).updateDeclExports(module, decl, exports),
             .c => return {},
+            .wasm => return @fieldParentPtr(Wasm, "base", base).updateDeclExports(module, decl, exports),
         }
     }
 
@@ -204,6 +218,7 @@ pub const File = struct {
         elf,
         macho,
         c,
+        wasm,
     };
 
     pub const ErrorFlags = struct {
@@ -2832,6 +2847,7 @@ pub const File = struct {
     };
 
     pub const MachO = @import("link/MachO.zig");
+    const Wasm = @import("link/Wasm.zig");
 };
 
 /// Saturating multiplication
diff --git a/src-self-hosted/link/Wasm.zig b/src-self-hosted/link/Wasm.zig
new file mode 100644
index 0000000000..a62cff15ee
--- /dev/null
+++ b/src-self-hosted/link/Wasm.zig
@@ -0,0 +1,445 @@
+const Wasm = @This();
+
+const std = @import("std");
+const Allocator = std.mem.Allocator;
+const assert = std.debug.assert;
+const fs = std.fs;
+const leb = std.debug.leb;
+
+const Module = @import("../Module.zig");
+const codegen = @import("../codegen/wasm.zig");
+const link = @import("../link.zig");
+
+/// Various magic numbers defined by the wasm spec
+const spec = struct {
+    const magic = [_]u8{ 0x00, 0x61, 0x73, 0x6D }; // \0asm
+    const version = [_]u8{ 0x01, 0x00, 0x00, 0x00 }; // version 1
+
+    const custom_id = 0;
+    const types_id = 1;
+    const imports_id = 2;
+    const funcs_id = 3;
+    const tables_id = 4;
+    const memories_id = 5;
+    const globals_id = 6;
+    const exports_id = 7;
+    const start_id = 8;
+    const elements_id = 9;
+    const code_id = 10;
+    const data_id = 11;
+};
+
+pub const base_tag = link.File.Tag.wasm;
+
+pub const FnData = struct {
+    funcidx: u32,
+    typeidx: u32,
+};
+
+base: link.File,
+
+types: Types,
+funcs: Funcs,
+exports: Exports,
+
+/// Array over the section structs used in the various sections above to
+/// allow iteration when shifting sections to make space.
+/// TODO: this should eventually be size 11 when we use all the sections.
+sections: [4]*Section,
+
+pub fn openPath(allocator: *Allocator, dir: fs.Dir, sub_path: []const u8, options: link.Options) !*link.File {
+    assert(options.object_format == .wasm);
+
+    // TODO: read the file and keep vaild parts instead of truncating
+    const file = try dir.createFile(sub_path, .{ .truncate = true, .read = true });
+    errdefer file.close();
+
+    const wasm = try allocator.create(Wasm);
+    errdefer allocator.destroy(wasm);
+
+    try file.writeAll(&(spec.magic ++ spec.version));
+
+    wasm.base = .{
+        .tag = .wasm,
+        .options = options,
+        .file = file,
+        .allocator = allocator,
+    };
+
+    // TODO: this should vary depending on the section and be less arbitrary
+    const size = 1024;
+    const offset = @sizeOf(@TypeOf(spec.magic ++ spec.version));
+
+    wasm.types = try Types.init(file, offset, size);
+    wasm.funcs = try Funcs.init(file, offset + size, size, offset + 3 * size, size);
+    wasm.exports = try Exports.init(file, offset + 2 * size, size);
+    try file.setEndPos(offset + 4 * size);
+
+    wasm.sections = [_]*Section{
+        &wasm.types.typesec.section,
+        &wasm.funcs.funcsec,
+        &wasm.exports.exportsec,
+        &wasm.funcs.codesec.section,
+    };
+
+    return &wasm.base;
+}
+
+pub fn deinit(self: *Wasm) void {
+    if (self.base.file) |f| f.close();
+    self.types.deinit();
+    self.funcs.deinit();
+}
+
+pub fn updateDecl(self: *Wasm, module: *Module, decl: *Module.Decl) !void {
+    if (decl.typed_value.most_recent.typed_value.ty.zigTypeTag() != .Fn)
+        return error.TODOImplementNonFnDeclsForWasm;
+
+    if (decl.fn_link.wasm) |fn_data| {
+        self.types.free(fn_data.typeidx);
+        self.funcs.free(fn_data.funcidx);
+    }
+
+    var buf = std.ArrayList(u8).init(self.base.allocator);
+    defer buf.deinit();
+
+    try codegen.genFunctype(&buf, decl);
+    const typeidx = try self.types.new(buf.items);
+    buf.items.len = 0;
+
+    try codegen.genCode(&buf, decl);
+    const funcidx = try self.funcs.new(typeidx, buf.items);
+
+    decl.fn_link.wasm = .{ .typeidx = typeidx, .funcidx = funcidx };
+
+    try self.exports.writeAll(module);
+}
+
+pub fn updateDeclExports(
+    self: *Wasm,
+    module: *Module,
+    decl: *const Module.Decl,
+    exports: []const *Module.Export,
+) !void {
+    // TODO: updateDeclExports() may currently be called before updateDecl,
+    // presumably due to a bug. For now just rely on the following call
+    // being made in updateDecl().
+
+    //try self.exports.writeAll(module);
+}
+
+pub fn freeDecl(self: *Wasm, decl: *Module.Decl) void {
+    // TODO: remove this assert when non-function Decls are implemented
+    assert(decl.typed_value.most_recent.typed_value.ty.zigTypeTag() == .Fn);
+    if (decl.fn_link.wasm) |fn_data| {
+        self.types.free(fn_data.typeidx);
+        self.funcs.free(fn_data.funcidx);
+        decl.fn_link.wasm = null;
+    }
+}
+
+pub fn flush(self: *Wasm) !void {}
+
+/// This struct describes the location of a named section + custom section
+/// padding in the output file. This is all the data we need to allow for
+/// shifting sections around when padding runs out.
+const Section = struct {
+    /// The size of a section header: 1 byte section id + 5 bytes
+    /// for the fixed-width ULEB128 encoded contents size.
+    const header_size = 1 + 5;
+    /// Offset of the section id byte from the start of the file.
+    offset: u64,
+    /// Size of the section, including the header and directly
+    /// following custom section used for padding if any.
+    size: u64,
+
+    /// Resize the usable part of the section, handling the following custom
+    /// section used for padding. If there is not enough padding left, shift
+    /// all following sections to make space. Takes the current and target
+    /// contents sizes of the section as arguments.
+    fn resize(self: *Section, file: fs.File, current: u32, target: u32) !void {
+        // Section header + target contents size + custom section header
+        // + custom section name + empty custom section > owned chunk of the file
+        if (header_size + target + header_size + 1 + 0 > self.size)
+            return error.TODOImplementSectionShifting;
+
+        const new_custom_start = self.offset + header_size + target;
+        const new_custom_contents_size = self.size - target - 2 * header_size;
+        assert(new_custom_contents_size >= 1);
+        // +1 for the name of the custom section, which we set to an empty string
+        var custom_header: [header_size + 1]u8 = undefined;
+        custom_header[0] = spec.custom_id;
+        leb.writeUnsignedFixed(5, custom_header[1..header_size], @intCast(u32, new_custom_contents_size));
+        custom_header[header_size] = 0;
+        try file.pwriteAll(&custom_header, new_custom_start);
+    }
+};
+
+/// This can be used to manage the contents of any section which uses a vector
+/// of contents. This interface maintains index stability while allowing for
+/// reuse of "dead" indexes.
+const VecSection = struct {
+    /// Represents a single entry in the vector (e.g. a type in the type section)
+    const Entry = struct {
+        /// Offset from the start of the section contents in bytes
+        offset: u32,
+        /// Size in bytes of the entry
+        size: u32,
+    };
+    section: Section,
+    /// Size in bytes of the contents of the section. Does not include
+    /// the "header" containing the section id and this value.
+    contents_size: u32,
+    /// List of all entries in the contents of the section.
+    entries: std.ArrayListUnmanaged(Entry) = std.ArrayListUnmanaged(Entry){},
+    /// List of indexes of unreferenced entries which may be
+    /// overwritten and reused.
+    dead_list: std.ArrayListUnmanaged(u32) = std.ArrayListUnmanaged(u32){},
+
+    /// Write the headers of the section and custom padding section
+    fn init(comptime section_id: u8, file: fs.File, offset: u64, initial_size: u64) !VecSection {
+        // section id, section size, empty vector, custom section id,
+        // custom section size, empty custom section name
+        var initial_data: [1 + 5 + 5 + 1 + 5 + 1]u8 = undefined;
+
+        assert(initial_size >= initial_data.len);
+
+        comptime var i = 0;
+        initial_data[i] = section_id;
+        i += 1;
+        leb.writeUnsignedFixed(5, initial_data[i..(i + 5)], 5);
+        i += 5;
+        leb.writeUnsignedFixed(5, initial_data[i..(i + 5)], 0);
+        i += 5;
+        initial_data[i] = spec.custom_id;
+        i += 1;
+        leb.writeUnsignedFixed(5, initial_data[i..(i + 5)], @intCast(u32, initial_size - @sizeOf(@TypeOf(initial_data))));
+        i += 5;
+        initial_data[i] = 0;
+
+        try file.pwriteAll(&initial_data, offset);
+
+        return VecSection{
+            .section = .{
+                .offset = offset,
+                .size = initial_size,
+            },
+            .contents_size = 5,
+        };
+    }
+
+    fn deinit(self: *VecSection, allocator: *Allocator) void {
+        self.entries.deinit(allocator);
+        self.dead_list.deinit(allocator);
+    }
+
+    /// Write a new entry into the file, returning the index used.
+    fn addEntry(self: *VecSection, file: fs.File, allocator: *Allocator, data: []const u8) !u32 {
+        // First look for a dead entry we can reuse
+        for (self.dead_list.items) |dead_idx, i| {
+            const dead_entry = &self.entries.items[dead_idx];
+            if (dead_entry.size == data.len) {
+                // Found a dead entry of the right length, overwrite it
+                try file.pwriteAll(data, self.section.offset + Section.header_size + dead_entry.offset);
+                _ = self.dead_list.swapRemove(i);
+                return dead_idx;
+            }
+        }
+
+        // TODO: We can be more efficient if we special-case one or
+        // more consecutive dead entries at the end of the vector.
+
+        // We failed to find a dead entry to reuse, so write the new
+        // entry to the end of the section.
+        try self.section.resize(file, self.contents_size, self.contents_size + @intCast(u32, data.len));
+        try file.pwriteAll(data, self.section.offset + Section.header_size + self.contents_size);
+        try self.entries.append(allocator, .{
+            .offset = self.contents_size,
+            .size = @intCast(u32, data.len),
+        });
+        self.contents_size += @intCast(u32, data.len);
+        // Make sure the dead list always has enough space to store all free'd
+        // entries. This makes it so that delEntry() cannot fail.
+        // TODO: figure out a better way that doesn't waste as much memory
+        try self.dead_list.ensureCapacity(allocator, self.entries.items.len);
+
+        // Update the size in the section header and the item count of
+        // the contents vector.
+        var size_and_count: [10]u8 = undefined;
+        leb.writeUnsignedFixed(5, size_and_count[0..5], self.contents_size);
+        leb.writeUnsignedFixed(5, size_and_count[5..], @intCast(u32, self.entries.items.len));
+        try file.pwriteAll(&size_and_count, self.section.offset + 1);
+
+        return @intCast(u32, self.entries.items.len - 1);
+    }
+
+    /// Mark the type referenced by the given index as dead.
+    fn delEntry(self: *VecSection, index: u32) void {
+        self.dead_list.appendAssumeCapacity(index);
+    }
+};
+
+const Types = struct {
+    typesec: VecSection,
+
+    fn init(file: fs.File, offset: u64, initial_size: u64) !Types {
+        return Types{ .typesec = try VecSection.init(spec.types_id, file, offset, initial_size) };
+    }
+
+    fn deinit(self: *Types) void {
+        const wasm = @fieldParentPtr(Wasm, "types", self);
+        self.typesec.deinit(wasm.base.allocator);
+    }
+
+    fn new(self: *Types, data: []const u8) !u32 {
+        const wasm = @fieldParentPtr(Wasm, "types", self);
+        return self.typesec.addEntry(wasm.base.file.?, wasm.base.allocator, data);
+    }
+
+    fn free(self: *Types, typeidx: u32) void {
+        self.typesec.delEntry(typeidx);
+    }
+};
+
+const Funcs = struct {
+    /// This section needs special handling to keep the indexes matching with
+    /// the codesec, so we cant just use a VecSection.
+    funcsec: Section,
+    /// Number of functions listed in the funcsec. Must be kept in sync with
+    /// codesec.entries.items.len.
+    funcs_count: u32,
+    codesec: VecSection,
+
+    fn init(file: fs.File, funcs_offset: u64, funcs_size: u64, code_offset: u64, code_size: u64) !Funcs {
+        return Funcs{
+            .funcsec = (try VecSection.init(spec.funcs_id, file, funcs_offset, funcs_size)).section,
+            .funcs_count = 0,
+            .codesec = try VecSection.init(spec.code_id, file, code_offset, code_size),
+        };
+    }
+
+    fn deinit(self: *Funcs) void {
+        const wasm = @fieldParentPtr(Wasm, "funcs", self);
+        self.codesec.deinit(wasm.base.allocator);
+    }
+
+    /// Add a new function to the binary, first finding space for and writing
+    /// the code then writing the typeidx to the corresponding index in the
+    /// funcsec. Returns the function index used.
+    fn new(self: *Funcs, typeidx: u32, code: []const u8) !u32 {
+        const wasm = @fieldParentPtr(Wasm, "funcs", self);
+        const file = wasm.base.file.?;
+        const allocator = wasm.base.allocator;
+
+        assert(self.funcs_count == self.codesec.entries.items.len);
+
+        // TODO: consider nop-padding the code if there is a close but not perfect fit
+        const funcidx = try self.codesec.addEntry(file, allocator, code);
+
+        if (self.funcs_count < self.codesec.entries.items.len) {
+            // u32 vector length + funcs_count u32s in the vector
+            const current = 5 + self.funcs_count * 5;
+            try self.funcsec.resize(file, current, current + 5);
+            self.funcs_count += 1;
+
+            // Update the size in the section header and the item count of
+            // the contents vector.
+            var size_and_count: [10]u8 = undefined;
+            leb.writeUnsignedFixed(5, size_and_count[0..5], 5 + self.funcs_count * 5);
+            leb.writeUnsignedFixed(5, size_and_count[5..], self.funcs_count);
+            try file.pwriteAll(&size_and_count, self.funcsec.offset + 1);
+        }
+        assert(self.funcs_count == self.codesec.entries.items.len);
+
+        var typeidx_leb: [5]u8 = undefined;
+        leb.writeUnsignedFixed(5, &typeidx_leb, typeidx);
+        try file.pwriteAll(&typeidx_leb, self.funcsec.offset + Section.header_size + 5 + funcidx * 5);
+
+        return funcidx;
+    }
+
+    fn free(self: *Funcs, funcidx: u32) void {
+        self.codesec.delEntry(funcidx);
+    }
+};
+
+/// Exports are tricky. We can't leave dead entries in the binary as they
+/// would obviously be visible from the execution environment. The simplest
+/// way to work around this is to re-emit the export section whenever
+/// something changes. This also makes it easier to ensure exported function
+/// and global indexes are updated as they change.
+const Exports = struct {
+    exportsec: Section,
+    /// Size in bytes of the contents of the section. Does not include
+    /// the "header" containing the section id and this value.
+    contents_size: u32,
+
+    fn init(file: fs.File, offset: u64, initial_size: u64) !Exports {
+        return Exports{
+            .exportsec = (try VecSection.init(spec.exports_id, file, offset, initial_size)).section,
+            .contents_size = 5,
+        };
+    }
+
+    fn writeAll(self: *Exports, module: *Module) !void {
+        const wasm = @fieldParentPtr(Wasm, "exports", self);
+        const file = wasm.base.file.?;
+        var buf: [5]u8 = undefined;
+
+        // First ensure the section is the right size
+        var export_count: u32 = 0;
+        var new_contents_size: u32 = 5;
+        for (module.decl_exports.entries.items) |entry| {
+            for (entry.value) |e| {
+                export_count += 1;
+                new_contents_size += calcSize(e);
+            }
+        }
+        if (new_contents_size != self.contents_size) {
+            try self.exportsec.resize(file, self.contents_size, new_contents_size);
+            leb.writeUnsignedFixed(5, &buf, new_contents_size);
+            try file.pwriteAll(&buf, self.exportsec.offset + 1);
+        }
+
+        try file.seekTo(self.exportsec.offset + Section.header_size);
+        const writer = file.writer();
+
+        // Length of the exports vec
+        leb.writeUnsignedFixed(5, &buf, export_count);
+        try writer.writeAll(&buf);
+
+        for (module.decl_exports.entries.items) |entry|
+            for (entry.value) |e| try writeExport(writer, e);
+    }
+
+    /// Return the total number of bytes an export will take.
+    /// TODO: fixed-width LEB128 is currently used for simplicity, but should
+    /// be replaced with proper variable-length LEB128 as it is inefficient.
+    fn calcSize(e: *Module.Export) u32 {
+        // LEB128 name length + name bytes + export type + LEB128 index
+        return 5 + @intCast(u32, e.options.name.len) + 1 + 5;
+    }
+
+    /// Write the data for a single export to the given file at a given offset.
+    /// TODO: fixed-width LEB128 is currently used for simplicity, but should
+    /// be replaced with proper variable-length LEB128 as it is inefficient.
+    fn writeExport(writer: anytype, e: *Module.Export) !void {
+        var buf: [5]u8 = undefined;
+
+        // Export name length + name
+        leb.writeUnsignedFixed(5, &buf, @intCast(u32, e.options.name.len));
+        try writer.writeAll(&buf);
+        try writer.writeAll(e.options.name);
+
+        switch (e.exported_decl.typed_value.most_recent.typed_value.ty.zigTypeTag()) {
+            .Fn => {
+                // Type of the export
+                try writer.writeByte(0x00);
+                // Exported function index
+                leb.writeUnsignedFixed(5, &buf, e.exported_decl.fn_link.wasm.?.funcidx);
+                try writer.writeAll(&buf);
+            },
+            else => return error.TODOImplementNonFnDeclsForWasm,
+        }
+    }
+};
diff --git a/src-self-hosted/main.zig b/src-self-hosted/main.zig
index d63ea3a757..52e6a3b651 100644
--- a/src-self-hosted/main.zig
+++ b/src-self-hosted/main.zig
@@ -150,6 +150,7 @@ const usage_build_generic =
     \\  -ofmt=[mode]              Override target object format
     \\    elf                     Executable and Linking Format
     \\    c                       Compile to C source code
+    \\    wasm                    WebAssembly
     \\    coff   (planned)        Common Object File Format (Windows)
     \\    pe     (planned)        Portable Executable (Windows)
     \\    macho  (planned)        macOS relocatables