diff --git a/src/network/cache/FsCache.zig b/src/network/cache/FsCache.zig
new file mode 100644
index 00000000..db916cfa
--- /dev/null
+++ b/src/network/cache/FsCache.zig
@@ -0,0 +1,239 @@
+// Copyright (C) 2023-2026  Lightpanda (Selecy SAS)
+//
+// Francis Bouvier <francis@lightpanda.io>
+// Pierre Tachoire <pierre@lightpanda.io>
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as
+// published by the Free Software Foundation, either version 3 of the
+// License, or (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+
+const std = @import("std");
+const Cache = @import("Cache.zig");
+const CachedMetadata = Cache.CachedMetadata;
+const CachedResponse = Cache.CachedResponse;
+
+pub const FsCache = @This();
+
+dir: std.fs.Dir,
+
+pub fn init(path: []const u8) !FsCache {
+    const cwd = std.fs.cwd();
+
+    cwd.makeDir(path) catch |err| switch (err) {
+        error.PathAlreadyExists => {},
+        else => return err,
+    };
+
+    const dir = try cwd.openDir(path, .{ .iterate = true });
+    return .{ .dir = dir };
+}
+
+pub fn deinit(self: *FsCache) void {
+    self.dir.close();
+}
+
+pub fn cache(self: *FsCache) Cache {
+    return Cache.init(self);
+}
+
+const HASHED_KEY_LEN = 16;
+const HASHED_PATH_LEN = HASHED_KEY_LEN + 5;
+const HASHED_TMP_PATH_LEN = HASHED_PATH_LEN + 4;
+
+fn hashKey(key: []const u8) [HASHED_KEY_LEN]u8 {
+    const h = std.hash.Wyhash.hash(0, key);
+    var hex: [HASHED_KEY_LEN]u8 = undefined;
+    _ = std.fmt.bufPrint(&hex, "{x:0>16}", .{h}) catch unreachable;
+    return hex;
+}
+
+fn serializeMeta(writer: *std.Io.Writer, meta: *const CachedMetadata) !void {
+    try writer.print("{s}\n{s}\n", .{ meta.url, meta.content_type });
+    try writer.print("{d}\n{d}\n{d}\n{d}\n", .{
+        meta.status,
+        meta.stored_at,
+        meta.age_at_store,
+        meta.max_age,
+    });
+    try writer.print("{s}\n", .{meta.etag orelse "null"});
+    try writer.print("{s}\n", .{meta.last_modified orelse "null"});
+    try writer.print("{s}\n", .{meta.vary orelse "null"});
+    try writer.print("{}\n{}\n{}\n", .{
+        meta.must_revalidate,
+        meta.no_cache,
+        meta.immutable,
+    });
+    try writer.flush();
+}
+
+fn deserializeMetaOptionalString(bytes: []const u8) ?[]const u8 {
+    if (std.mem.eql(u8, bytes, "null")) return null else return bytes;
+}
+
+fn deserializeMetaBoolean(bytes: []const u8) !bool {
+    if (std.mem.eql(u8, bytes, "true")) return true;
+    if (std.mem.eql(u8, bytes, "false")) return false;
+    return error.Malformed;
+}
+
+fn deserializeMeta(allocator: std.mem.Allocator, file: std.fs.File) !CachedMetadata {
+    var file_buf: [1024]u8 = undefined;
+    var file_reader = file.reader(&file_buf);
+    const reader = &file_reader.interface;
+
+    const url = blk: {
+        const line = try reader.takeDelimiter('\n') orelse return error.Malformed;
+        break :blk try allocator.dupeZ(u8, line);
+    };
+    errdefer allocator.free(url);
+
+    const content_type = blk: {
+        const line = try reader.takeDelimiter('\n') orelse return error.Malformed;
+        break :blk try allocator.dupe(u8, line);
+    };
+    errdefer allocator.free(content_type);
+
+    const status = blk: {
+        const line = try reader.takeDelimiter('\n') orelse return error.Malformed;
+        break :blk std.fmt.parseInt(u16, line, 10) catch return error.Malformed;
+    };
+    const stored_at = blk: {
+        const line = try reader.takeDelimiter('\n') orelse return error.Malformed;
+        break :blk std.fmt.parseInt(i64, line, 10) catch return error.Malformed;
+    };
+    const age_at_store = blk: {
+        const line = try reader.takeDelimiter('\n') orelse return error.Malformed;
+        break :blk std.fmt.parseInt(u64, line, 10) catch return error.Malformed;
+    };
+    const max_age = blk: {
+        const line = try reader.takeDelimiter('\n') orelse return error.Malformed;
+        break :blk std.fmt.parseInt(u64, line, 10) catch return error.Malformed;
+    };
+
+    const etag = blk: {
+        const line = try reader.takeDelimiter('\n') orelse return error.Malformed;
+        break :blk if (std.mem.eql(u8, line, "null")) null else try allocator.dupe(u8, line);
+    };
+    const last_modified = blk: {
+        const line = try reader.takeDelimiter('\n') orelse return error.Malformed;
+        break :blk if (std.mem.eql(u8, line, "null")) null else try allocator.dupe(u8, line);
+    };
+    const vary = blk: {
+        const line = try reader.takeDelimiter('\n') orelse return error.Malformed;
+        break :blk if (std.mem.eql(u8, line, "null")) null else try allocator.dupe(u8, line);
+    };
+
+    const must_revalidate = blk: {
+        const line = try reader.takeDelimiter('\n') orelse return error.Malformed;
+        break :blk try deserializeMetaBoolean(line);
+    };
+    const no_cache = blk: {
+        const line = try reader.takeDelimiter('\n') orelse return error.Malformed;
+        break :blk try deserializeMetaBoolean(line);
+    };
+    const immutable = blk: {
+        const line = try reader.takeDelimiter('\n') orelse return error.Malformed;
+        break :blk try deserializeMetaBoolean(line);
+    };
+
+    return .{
+        .url = url,
+        .content_type = content_type,
+        .status = status,
+        .stored_at = stored_at,
+        .age_at_store = age_at_store,
+        .max_age = max_age,
+        .etag = etag,
+        .last_modified = last_modified,
+        .must_revalidate = must_revalidate,
+        .no_cache = no_cache,
+        .immutable = immutable,
+        .vary = vary,
+    };
+}
+
+pub fn get(self: *FsCache, allocator: std.mem.Allocator, key: []const u8) ?Cache.CachedResponse {
+    const hashed_key = hashKey(key);
+
+    var meta_path: [HASHED_PATH_LEN]u8 = undefined;
+    _ = std.fmt.bufPrint(&meta_path, "{s}.meta", .{hashed_key}) catch @panic("FsCache.get meta path overflowed");
+
+    var body_path: [HASHED_PATH_LEN]u8 = undefined;
+    _ = std.fmt.bufPrint(&body_path, "{s}.body", .{hashed_key}) catch @panic("FsCache.get body path overflowed");
+
+    const meta_file = self.dir.openFile(&meta_path, .{ .mode = .read_only }) catch return null;
+    defer meta_file.close();
+
+    const meta = deserializeMeta(allocator, meta_file) catch {
+        self.dir.deleteFile(&meta_path) catch {};
+        self.dir.deleteFile(&body_path) catch {};
+        return null;
+    };
+
+    const body_file = self.dir.openFile(&body_path, .{ .mode = .read_only }) catch return null;
+
+    return .{
+        .metadata = meta,
+        .data = .{ .file = body_file },
+    };
+}
+
+pub fn put(self: *FsCache, key: []const u8, meta: CachedMetadata, body: []const u8) !void {
+    const hashed_key = hashKey(key);
+
+    // Write meta to a temp file, then atomically rename into place
+    var meta_path: [HASHED_PATH_LEN]u8 = undefined;
+    _ = std.fmt.bufPrint(&meta_path, "{s}.meta", .{hashed_key}) catch
+        @panic("FsCache.put meta path overflowed");
+
+    var meta_tmp_path: [HASHED_TMP_PATH_LEN]u8 = undefined;
+    _ = std.fmt.bufPrint(&meta_tmp_path, "{s}.meta.tmp", .{hashed_key}) catch
+        @panic("FsCache.put meta tmp path overflowed");
+
+    {
+        const meta_file = try self.dir.createFile(&meta_tmp_path, .{});
+        errdefer {
+            meta_file.close();
+            self.dir.deleteFile(&meta_tmp_path) catch {};
+        }
+
+        var buf: [512]u8 = undefined;
+        var meta_file_writer = meta_file.writer(&buf);
+        try serializeMeta(&meta_file_writer.interface, &meta);
+        meta_file.close();
+    }
+    errdefer self.dir.deleteFile(&meta_tmp_path) catch {};
+    try self.dir.rename(&meta_tmp_path, &meta_path);
+
+    // Write body to a temp file, then atomically rename into place
+    var body_path: [HASHED_PATH_LEN]u8 = undefined;
+    _ = std.fmt.bufPrint(&body_path, "{s}.body", .{hashed_key}) catch
+        @panic("FsCache.put body path overflowed");
+
+    var body_tmp_path: [HASHED_TMP_PATH_LEN]u8 = undefined;
+    _ = std.fmt.bufPrint(&body_tmp_path, "{s}.body.tmp", .{hashed_key}) catch
+        @panic("FsCache.put body tmp path overflowed");
+
+    {
+        const body_file = try self.dir.createFile(&body_tmp_path, .{});
+        errdefer {
+            body_file.close();
+            self.dir.deleteFile(&body_tmp_path) catch {};
+        }
+        try body_file.writeAll(body);
+        body_file.close();
+    }
+    errdefer self.dir.deleteFile(&body_tmp_path) catch {};
+
+    errdefer self.dir.deleteFile(&meta_path) catch {};
+    try self.dir.rename(&body_tmp_path, &body_path);
+}