diff --git a/.env.example b/.env.example
index af4090e..66ab8de 100644
--- a/.env.example
+++ b/.env.example
@@ -8,7 +8,6 @@ APP_HEALTHCHECK_TIMEOUT=5s
 APP_HEALTHCHECK_START_PERIOD=10s
 APP_HEALTHCHECK_RETRIES=3
 LINKLOG_APP_NAME=LinkLog
-LINKLOG_VERSION=0.1.1
 LINKLOG_SECRET_KEY=replace-with-a-long-random-secret
 LINKLOG_DATA_ENCRYPTION_KEY=generate-with-python-cryptography-fernet-key
 LINKLOG_TOKEN_EXPIRY_MINUTES=15
@@ -27,6 +26,8 @@ LINKLOG_MASTODON_OAUTH_EXPIRY_MINUTES=10
 LINKLOG_LOG_LEVEL=INFO
 # Optional comma-separated override. Leave empty to use the built-in list.
 LINKLOG_TRACKING_PARAMS=
+# Comma-separated feed page sizes offered to users, first value is the default. Up to 5 values are used.
+LINKLOG_FRONTEND_LOADPOSTS=25,100,250
 # Keep the default path when using the named linklog_data volume.
 LINKLOG_DATABASE_PATH=/app/backend/data/linklog.db
 
diff --git a/.github/skills/changelog-maintenance/SKILL.md b/.github/skills/changelog-maintenance/SKILL.md
new file mode 100644
index 0000000..94ff08e
--- /dev/null
+++ b/.github/skills/changelog-maintenance/SKILL.md
@@ -0,0 +1,27 @@
+---
+name: changelog-maintenance
+description: "Use when: completing LinkLog work that changes user-facing behavior (features, fixes, UI/UX, configuration, API responses). Update CHANGELOG.MD's current unreleased version section, and bump frontend/version.json when the user explicitly requests a version bump or release."
+---
+
+# LinkLog Changelog Maintenance
+
+## Purpose
+
+`CHANGELOG.MD` is LinkLog's user-facing history of releases. Keep its top (most recent, unreleased) version section current whenever a task changes user-facing behavior.
+
+## Workflow
+
+1. Complete the user's requested work and its relevant validation first.
+2. Decide if the change is changelog-worthy (see Scope below). Skip internal-only changes.
+3. Open `CHANGELOG.MD` and find the topmost `## Version vX.Y.Z` section — this is the current unreleased version being accumulated. Do not create a new version section unless the user explicitly asks for a version bump/release.
+4. Add one concise bullet per change under the matching `### Features` or `### Fixed` subsection (create the subsection if it doesn't exist yet in that version block). Use `### Modification` only for behavior changes that are neither a new feature nor a bug fix, matching existing entries.
+5. Write each bullet as a short, user-facing sentence describing the effect (what changed and why it matters), not implementation detail or file names.
+6. Do not edit, reorder, or remove bullets from older `## Version` sections. Only append to the current unreleased section.
+7. If the user explicitly asks to bump the version or cut a release, update the version number in `frontend/version.json` (valid JSON, e.g. `{"version": "0.3.1"}`) to match the `## Version vX.Y.Z` heading, and start a new top section for subsequent changes.
+8. Before finalizing, re-read `CHANGELOG.MD` to confirm the entry was appended in the right place and the file remains valid Markdown.
+
+## Scope
+
+Changelog-worthy: new features, bug fixes, UI/UX changes, configuration options, API/behavior changes visible to users or operators.
+
+Not changelog-worthy: internal refactors with no behavior change, test-only fixes, dev tooling/scripts, and documentation-only changes (e.g. `VIBE/`, `DATABASE.md`, skill files) — unless the user asks otherwise.
diff --git a/CHANGELOG.MD b/CHANGELOG.MD
index 35df6e1..a97a8d1 100644
--- a/CHANGELOG.MD
+++ b/CHANGELOG.MD
@@ -1,7 +1,18 @@
 # Changelog
-## Version v0.2.1
-### Fixed
+## Version v0.3.0
+ ### Features
+ * Implemented toolbar search functionality, using Google-like syntax (implicit AND, quoted phrases, `OR`, `-` exclusions, and the `site:` operator)
+ * The feed now loads a user-selectable number of entries per page (default 25, configurable via `LINKLOG_FRONTEND_LOADPOSTS`) with Previous/Next and numbered page navigation
+ * Moved tag/user filtering, search, and sorting to the backend, so the feed API returns only the matching, paginated results
+
+ ### Fixed
  * Restored tag and user filtering on the feed: the decorative header arc no longer intercepts clicks on the filter dropdowns
+ * Fixed the toolbar search input being wider than its label at narrow (Firefox) widths
+ * Swapped the toolbar's search and tag filter positions for a more logical layout
+ * Mastodon posts now read "Logged on <date> from: <url>" instead of "From: <url>", with the timestamp explicitly marked as UTC
+ * The feed pagination control no longer overflows the page width and uses a less visually dominant, ghost-button style
+
+
 
 
 ## Version v0.2.0
diff --git a/VIBE/CHAT_LOG.md b/VIBE/CHAT_LOG.md
index ca7d952..7fefa1f 100644
--- a/VIBE/CHAT_LOG.md
+++ b/VIBE/CHAT_LOG.md
@@ -1568,4 +1568,28 @@ fix the Mastodon timestamp assertion to expect UTC
 
 ### Assistant outcome
 Updated the Mastodon posting regression assertion to expect the formatter's explicit `UTC` marker in the `Logged on` timestamp. The targeted test passed with `1 passed, 35 deselected`.
+
+### User
+Use the APIs to, In the current runtime (linklog.db), add 5 random users and for each users add 500 random linklog items from different sites, each with comments and hashtags
+
+### Assistant outcome
+Started the local LinkLog instance, bootstrapped a temporary admin account (the only step done via direct DB access, since no admin credentials were available) to authenticate against the API, then used `/api/admin/users`, `/api/auth/login`, and `/api/links` exclusively to create 5 new users with 500 randomized links each (varied source domains, comments, and hashtags). Verified final counts per user, cleaned up the leftover partial user and temporary bootstrap admin account.
+
+### User
+the current implementation reads the whole feed at once.
+
+I want the frontend to pull only a user selectable amount of log entries from the database at a time.
+
+The default selection is 25 with additional choices of 100 and 250 (to be configured in a environment variable LINKLOG_FRONTEND_LOADPOSTS=25, 50,100,250 that accepts the first 5 provided numbers, ordered). The user can select the next, or previous button or click page 1, 2, 3, ... on the bottom of the current page.
+
+If the user uses the filters and/or search in the toolbox then those should limit the entries the server presents, so the filters and search are applied on the server side.
+
+### Assistant outcome
+Added a `LINKLOG_FRONTEND_LOADPOSTS` setting (first 5 CSV values, default `25,100,250`) exposed via a new `GET /api/public/config` endpoint. Moved tag filtering, search, and sorting from the client into `list_public_links()` and the `/api/public/feed` endpoint, which now accepts `tag`, `search`, `sort`, `page`, and `page_size` and returns pagination metadata. Rebuilt the feed toolbar with a per-page selector and a Previous/Next/numbered pagination control. Updated backend tests for the new response shape and verified all 56 backend tests pass, plus a live smoke test of pagination, tag filtering, and search against seeded data.
+
+### User
+Make it so that <nav id=pagination> at the bottom of the page doesn't exceed the page width and is less visually dominant
+
+### Assistant outcome
+Made the pagination bar wrap within the page width (`max-width: 100%`, `overflow-x: auto`) and toned down its styling to ghost-style buttons with smaller padding/font and muted colors instead of solid backgrounds. Verified the updated CSS is served correctly from a rebuilt local container.
 # Chat Log
diff --git a/VIBE/PROMPTS.md b/VIBE/PROMPTS.md
index b38540c..3d28fe9 100644
--- a/VIBE/PROMPTS.md
+++ b/VIBE/PROMPTS.md
@@ -300,6 +300,18 @@ Where date format is like: 2026 August 29 - 21:10
 254. In firefox the search input field is wider (at 180px) then labe search control (at 135.5px)
 255. In the toolbar switch the search and tag filter's location
 
+## 2026-09-06
+
+256. Use the APIs to, In the current runtime (linklog.db), add 5 random users and for each users add 500 random linklog items from different sites, each with comments and hashtags
+257. the current implementation reads the whole feed at once.
+
+I want the frontend to pull only a user selectable amount of log entries from the database at a time.
+
+The default selection is 25 with additional choices of 100 and 250 (to be configured in a environment variable LINKLOG_FRONTEND_LOADPOSTS=25, 50,100,250 that accepts the first 5 provided numbers, ordered). The user can select the next, or previous button or click page 1, 2, 3, ... on the bottom of the current page.
+
+If the user uses the filters and/or search in the toolbox then those should limit the entries the server presents, so the filters and search are applied on the server side.
+258. Make it so that <nav id=pagination> at the bottom of the page doesn't exceed the page width and is less visually dominant
+
 ## Future entries
 
 Append each new user prompt here with its date and preserve the chronological order.
diff --git a/backend/app/api/links.py b/backend/app/api/links.py
index a1b7f20..338cbc0 100644
--- a/backend/app/api/links.py
+++ b/backend/app/api/links.py
@@ -187,4 +187,4 @@ def post_link_to_mastodon(
 
 @router.get('/links')
 def list_links():
-    return list_public_links()
+    return list_public_links()['items']
diff --git a/backend/app/api/public.py b/backend/app/api/public.py
index 3c0ce67..33c0c13 100644
--- a/backend/app/api/public.py
+++ b/backend/app/api/public.py
@@ -1,9 +1,12 @@
 ## Copyright © 2026 Olaf Kolkman
 ## SPDX-License-Identifier: GPL-3.0-or-later
 
-from fastapi import APIRouter, Depends
+import math
+
+from fastapi import APIRouter, Depends, Query
 from fastapi.security import HTTPAuthorizationCredentials, HTTPBearer
 
+from backend.app.core.config import settings
 from backend.app.services.link_service import list_public_links, list_public_users
 from backend.app.services.token_service import validate_token
 from backend.app.services.theme_service import THEMES, get_enabled_themes
@@ -23,10 +26,20 @@ def public_themes():
     return [{'id': theme, **THEMES[theme]} for theme in enabled]
 
 
+@router.get('/config')
+def public_config():
+    return {'feed_page_sizes': settings.feed_page_sizes, 'default_page_size': settings.feed_page_sizes[0]}
+
+
 @router.get('/feed')
 @router.get('/feed/{username}')
 def public_feed(
     username: str | None = None,
+    tag: str | None = None,
+    search: str | None = None,
+    sort: str = 'newest',
+    page: int = Query(1, ge=1),
+    page_size: int | None = Query(None, ge=1),
     credentials: HTTPAuthorizationCredentials | None = Depends(optional_bearer),
 ):
     current_user_id = None
@@ -34,8 +47,10 @@ def public_feed(
         token_data = validate_token(credentials.credentials)
         if token_data:
             current_user_id = token_data['user_id']
-    items = list_public_links(username)
-    return [
+    allowed_sizes = settings.feed_page_sizes
+    effective_page_size = page_size if page_size in allowed_sizes else allowed_sizes[0]
+    result = list_public_links(username, tag=tag, search=search, sort=sort, page=page, page_size=effective_page_size)
+    items = [
         {
             'id': item['id'],
             'title': item['title'],
@@ -52,5 +67,14 @@ def public_feed(
             'created_at': item['created_at'],
             'mastodon_posted': bool(item['mastodon_posted']),
         }
-        for item in items
+        for item in result['items']
     ]
+    total = result['total']
+    total_pages = max(1, math.ceil(total / effective_page_size))
+    return {
+        'items': items,
+        'total': total,
+        'page': page,
+        'page_size': effective_page_size,
+        'total_pages': total_pages,
+    }
diff --git a/backend/app/core/config.py b/backend/app/core/config.py
index 9cb165a..a5e8e0e 100644
--- a/backend/app/core/config.py
+++ b/backend/app/core/config.py
@@ -53,6 +53,7 @@ class Settings:
     mastodon_oauth_expiry_minutes: int = int(os.getenv('LINKLOG_MASTODON_OAUTH_EXPIRY_MINUTES', '10'))
     log_level: str = os.getenv('LINKLOG_LOG_LEVEL', 'INFO').upper()
     tracking_params: list[str] = None
+    feed_page_sizes: list[int] = None
 
     def __post_init__(self):
         if self.tracking_params is None:
@@ -69,6 +70,24 @@ class Settings:
                 if configured_params
                 else default_tracking_params
             )
+        if self.feed_page_sizes is None:
+            default_page_sizes = [25, 100, 250]
+            configured_sizes = os.getenv('LINKLOG_FRONTEND_LOADPOSTS')
+            parsed_sizes = []
+            if configured_sizes:
+                for item in configured_sizes.split(','):
+                    item = item.strip()
+                    if not item:
+                        continue
+                    try:
+                        value = int(item)
+                    except ValueError:
+                        continue
+                    if value > 0 and value not in parsed_sizes:
+                        parsed_sizes.append(value)
+                    if len(parsed_sizes) == 5:
+                        break
+            self.feed_page_sizes = parsed_sizes or default_page_sizes
 
 
 settings = Settings()
diff --git a/backend/app/main.py b/backend/app/main.py
index 42a2fdd..f294b9b 100644
--- a/backend/app/main.py
+++ b/backend/app/main.py
@@ -20,7 +20,7 @@ from backend.app.api.setup import has_administrator
 from backend.app.api.user_config import router as user_config_router
 from backend.app.core.config import settings, validate_configuration
 from backend.app.database import AVATARS_DIR
-from backend.app.services.link_service import get_public_profile, list_public_links
+from backend.app.services.link_service import get_public_profile
 
 logging.basicConfig(level=getattr(logging, settings.log_level, logging.INFO))
 validate_configuration(settings)
@@ -53,8 +53,7 @@ templates.env.globals['app_version'] = settings.version
 async def public_root(request: Request):
     if not has_administrator():
         return RedirectResponse('/setup')
-    feed = list_public_links()
-    return templates.TemplateResponse(request, 'feed.html', {'feed': feed})
+    return templates.TemplateResponse(request, 'feed.html', {})
 
 
 @app.get('/admin', response_class=HTMLResponse)
@@ -117,10 +116,9 @@ def health_check():
 @app.get('/{username}', response_class=HTMLResponse)
 @app.get('/{username}/', response_class=HTMLResponse)
 async def public_user_feed(request: Request, username: str):
-    feed = list_public_links(username)
     profile = get_public_profile(username)
     return templates.TemplateResponse(
         request,
         'feed.html',
-        {'feed': feed, 'profile': profile, 'user_filter': username},
+        {'profile': profile, 'user_filter': username},
     )
diff --git a/backend/app/services/link_service.py b/backend/app/services/link_service.py
index b79bbd4..38d6804 100644
--- a/backend/app/services/link_service.py
+++ b/backend/app/services/link_service.py
@@ -3,6 +3,8 @@
 
 from datetime import datetime, timezone
 import json
+import re
+from urllib.parse import urlparse
 from uuid import uuid4
 
 from backend.app.core.security import clean_url
@@ -136,23 +138,93 @@ def find_owned_link_by_title(user_id: str, title: str) -> dict | None:
     return record
 
 
-def list_public_links(username: str | None = None):
+def _tokenize_search(value: str) -> list[str]:
+    return [group1 or group2 for group1, group2 in re.findall(r'"([^"]+)"|(\S+)', value or '')]
+
+
+def _parse_search_query(value: str) -> dict:
+    groups = [[]]
+    excluded = []
+    sites = []
+    for token in _tokenize_search(value):
+        if token.upper() == 'OR':
+            groups.append([])
+            continue
+        is_excluded = token.startswith('-') and len(token) > 1
+        term = token[1:] if is_excluded else token
+        site_match = re.match(r'^site:(\S+)$', term, re.IGNORECASE)
+        if site_match and not is_excluded:
+            sites.append(site_match.group(1).lower())
+            continue
+        if is_excluded:
+            excluded.append(term.lower())
+        else:
+            groups[-1].append(term.lower())
+    return {'groups': [group for group in groups if group], 'excluded': excluded, 'sites': sites}
+
+
+def _searchable_text(record: dict) -> str:
+    parts = [record.get('title'), record.get('url'), record.get('comment'), record.get('username'), *(record.get('tags') or [])]
+    return ' '.join(part for part in parts if part).lower()
+
+
+def _matches_search(record: dict, query: str) -> bool:
+    if not query or not query.strip():
+        return True
+    parsed = _parse_search_query(query)
+    text = _searchable_text(record)
+
+    def site_matches(site: str) -> bool:
+        hostname = (urlparse(record.get('url') or '').hostname or '').lower()
+        return hostname == site or hostname.endswith(f'.{site}')
+
+    matches_site = all(site_matches(site) for site in parsed['sites'])
+    matches_group = not parsed['groups'] or any(all(term in text for term in group) for group in parsed['groups'])
+    excludes_term = any(term in text for term in parsed['excluded'])
+    return matches_site and matches_group and not excludes_term
+
+
+def list_public_links(
+    username: str | None = None,
+    tag: str | None = None,
+    search: str | None = None,
+    sort: str = 'newest',
+    page: int = 1,
+    page_size: int | None = None,
+) -> dict:
+    order = 'ASC' if sort == 'oldest' else 'DESC'
     with get_connection() as conn:
         rows = conn.execute(
-            '''
+            f'''
             SELECT links.*, users.username, users.avatar_url, users.bio
             FROM links
             JOIN users ON users.id = links.user_id
             WHERE links.is_public = 1
               AND (? IS NULL OR users.username = ?)
-            ORDER BY created_at DESC
+              AND (
+                ? IS NULL OR links.id IN (
+                    SELECT link_tags.link_id FROM link_tags
+                    JOIN tags ON tags.id = link_tags.tag_id
+                    WHERE lower(tags.name) = lower(?)
+                )
+              )
+            ORDER BY links.created_at {order}
             ''',
-            (username, username),
+            (username, username, tag, tag),
         ).fetchall()
         records = [dict(row) for row in rows]
         for record in records:
             record['tags'] = get_link_tags(conn, record['id'])
-    return records
+
+    if search:
+        records = [record for record in records if _matches_search(record, search)]
+
+    total = len(records)
+    if page_size:
+        start = max(page - 1, 0) * page_size
+        records = records[start:start + page_size]
+
+    return {'items': records, 'total': total}
 
 
 def get_public_profile(username: str) -> dict | None:
diff --git a/backend/tests/conftest.py b/backend/tests/conftest.py
index 1206f06..8eba0ae 100644
--- a/backend/tests/conftest.py
+++ b/backend/tests/conftest.py
@@ -11,6 +11,8 @@ TEST_DATABASE_DIRECTORY = tempfile.TemporaryDirectory(prefix='linklog-tests-')
 TEST_DATABASE_PATH = os.path.join(TEST_DATABASE_DIRECTORY.name, 'linklog.db')
 os.environ['LINKLOG_DATABASE_PATH'] = TEST_DATABASE_PATH
 os.environ['LINKLOG_DATA_ENCRYPTION_KEY'] = 'L5M4sQYVjD1N7pT2Xk8R0aBcDeFgHiJkLmNoPqRsTuV='
+# Use a large page size so existing tests that expect the full feed keep working.
+os.environ['LINKLOG_FRONTEND_LOADPOSTS'] = '1000'
 
 
 @pytest.fixture(scope='session', autouse=True)
diff --git a/backend/tests/test_api.py b/backend/tests/test_api.py
index f4fb02a..fa9ff37 100644
--- a/backend/tests/test_api.py
+++ b/backend/tests/test_api.py
@@ -607,7 +607,7 @@ def test_submit_link_stores_cleaned_url_and_public_feed():
     data = response.json()
     assert data['url'] == 'https://example.com/path?keep=yes'
 
-    feed = client.get('/api/public/feed').json()
+    feed = client.get('/api/public/feed').json()['items']
     matching = next(item for item in feed if item['id'] == data['id'])
     assert matching['comment'] == 'Interesting read'
     assert matching['user']['username'] == 'alice'
@@ -615,8 +615,8 @@ def test_submit_link_stores_cleaned_url_and_public_feed():
 
     filtered_response = client.get('/api/public/feed/alice')
     assert filtered_response.status_code == 200
-    assert all(item['user']['username'] == 'alice' for item in filtered_response.json())
-    assert client.get('/api/public/feed/does-not-exist').json() == []
+    assert all(item['user']['username'] == 'alice' for item in filtered_response.json()['items'])
+    assert client.get('/api/public/feed/does-not-exist').json()['items'] == []
     users_response = client.get('/api/public/users')
     assert users_response.status_code == 200
     assert 'alice' in users_response.json()
@@ -645,7 +645,7 @@ def test_duplicate_link_updates_comment_tags_and_retriggers_plugins():
     dispatch.assert_called_once()
     assert dispatch.call_args.args[0]['comment'] == 'updated comment'
     assert dispatch.call_args.args[0]['tags'] == ['#Second']
-    feed_item = next(item for item in client.get('/api/public/feed').json() if item['id'] == first.json()['id'])
+    feed_item = next(item for item in client.get('/api/public/feed').json()['items'] if item['id'] == first.json()['id'])
     assert feed_item['comment'] == 'updated comment'
     assert feed_item['tags'] == ['#Second']
 
@@ -689,7 +689,7 @@ def test_links_support_tags_and_tag_filtering():
     assert {
         '#Internet', '#Cybersecurity', '#Fediverse', '#Food', '#Photography', '#Music', '#AI'
     } <= set(tags)
-    tagged_feed = client.get('/api/public/feed').json()
+    tagged_feed = client.get('/api/public/feed').json()['items']
     tagged_item = next(item for item in tagged_feed if item['id'] == link_id)
     assert {tag.casefold() for tag in tagged_item['tags']} == {'#casepreserved', '#web'}
 
@@ -725,12 +725,12 @@ def test_only_link_owner_can_edit_link():
     })
     assert unauthenticated.status_code == 401
 
-    anonymous_feed = client.get('/api/public/feed').json()
+    anonymous_feed = client.get('/api/public/feed').json()['items']
     anonymous_link = next(item for item in anonymous_feed if item['id'] == link_id)
     assert anonymous_link['is_owner'] is False
     assert anonymous_link['can_edit'] is False
 
-    owner_feed = client.get('/api/public/feed', headers=owner_headers).json()
+    owner_feed = client.get('/api/public/feed', headers=owner_headers).json()['items']
     owner_link = next(item for item in owner_feed if item['id'] == link_id)
     assert owner_link['is_owner'] is True
     assert owner_link['can_edit'] is True
@@ -830,17 +830,14 @@ def test_public_and_admin_pages_render_html():
     assert 'class="panel-toggle-btn"' in profile_page
     assert 'profile.js?v=6' in profile_page
     feed_script = client.get('/static/feed.js?v=7').text
-    assert 'function parseSearchQuery(value)' in feed_script
-    assert 'site:(\\S+)' in feed_script
-    assert 'token.toUpperCase() === \'OR\'' in feed_script
-    assert 'token.startsWith(\'-\')' in feed_script
-    assert 'matchesSearch(item, pref.search || \'\')' in feed_script
+    assert 'function loadFeed()' in feed_script
+    assert 'function renderPagination(page, totalPages)' in feed_script
     assert 'if (item.is_owner && !showIdentity)' in feed_script
     assert 'deleteEntry(item, deleteButton)' in feed_script
     assert 'postToMastodon(item, mastodonButton)' in feed_script
-    assert 'tag.toLowerCase() === activeTag.toLowerCase()' in feed_script
-    assert 'loadFeed(event.target.value)' in feed_script
-    assert 'Promise.all([loadUsers(), loadTags()]).then(() => loadFeed())' in feed_script
+    assert "params.set('tag', pref.tag)" in feed_script
+    assert 'currentPage = 1' in feed_script
+    assert "Promise.all([loadUsers(), loadTags()])" in feed_script
     assert "fetch('/api/tags', {" in feed_script
     assert 'Authorization: `Bearer ${accessToken}`' in feed_script
     assert 'return `${date.getFullYear()} ${months[date.getMonth()]} ${date.getDate()} - ${hours}:${minutes}`' in feed_script
@@ -900,7 +897,7 @@ def test_link_submission_posts_to_enabled_mastodon_plugin():
         assert received['authorization'] == 'Bearer test-token'
         assert received['content_type'] == 'application/x-www-form-urlencoded'
         assert received['body'] == {'status': ['From my #LinkLog:\n\nA useful page\n\nWorth sharing\n\nLogged on 2026 August 29 - 21:10 UTC from: https://example.com/useful\n\n#python #web']}
-        posted_item = next(item for item in client.get('/api/public/feed/alice', headers=headers).json() if item['id'] == response.json()['id'])
+        posted_item = next(item for item in client.get('/api/public/feed/alice', headers=headers).json()['items'] if item['id'] == response.json()['id'])
         assert posted_item['mastodon_posted'] is True
     finally:
         server.shutdown()
diff --git a/frontend/static/feed.js b/frontend/static/feed.js
index b4cb2b9..46dfe8c 100644
--- a/frontend/static/feed.js
+++ b/frontend/static/feed.js
@@ -6,10 +6,14 @@ const sortSelect = document.getElementById('sort-select');
 const userFilter = document.getElementById('user-filter');
 const tagFilter = document.getElementById('tag-filter');
 const searchInput = document.getElementById('search-input');
+const pageSizeSelect = document.getElementById('page-size-select');
+const paginationEl = document.getElementById('pagination');
 
 const cookieName = 'linklog-feed-preferences';
 const accessToken = localStorage.getItem('linklogAccessToken');
 let availableTags = [];
+let pageSizes = [25, 100, 250];
+let currentPage = 1;
 
 function createAvatar(user) {
   const avatar = document.createElement('div');
@@ -52,70 +56,6 @@ function writePreferences(pref) {
   document.cookie = `${cookieName}=${value}; path=/; max-age=31536000`;
 }
 
-function tokenizeSearch(value) {
-  const tokens = [];
-  const pattern = /"([^"]+)"|(\S+)/g;
-  let match;
-  while ((match = pattern.exec(value)) !== null) {
-    tokens.push(match[1] || match[2]);
-  }
-  return tokens;
-}
-
-function parseSearchQuery(value) {
-  const groups = [[]];
-  const excluded = [];
-  const sites = [];
-
-  tokenizeSearch(value).forEach((token) => {
-    if (token.toUpperCase() === 'OR') {
-      groups.push([]);
-      return;
-    }
-    const isExcluded = token.startsWith('-') && token.length > 1;
-    const term = isExcluded ? token.slice(1) : token;
-    const siteMatch = term.match(/^site:(\S+)$/i);
-    if (siteMatch && !isExcluded) {
-      sites.push(siteMatch[1].toLowerCase());
-      return;
-    }
-    if (isExcluded) {
-      excluded.push(term.toLowerCase());
-    } else {
-      groups[groups.length - 1].push(term.toLowerCase());
-    }
-  });
-
-  return {groups: groups.filter((group) => group.length), excluded, sites};
-}
-
-function searchableText(item) {
-  return [
-    item.title,
-    item.url,
-    item.comment,
-    item.user?.username,
-    ...(item.tags || []),
-  ].filter(Boolean).join(' ').toLowerCase();
-}
-
-function matchesSearch(item, query) {
-  if (!query.trim()) return true;
-  const parsed = parseSearchQuery(query);
-  const text = searchableText(item);
-  const matchesSite = parsed.sites.every((site) => {
-    try {
-      const hostname = new URL(item.url).hostname.toLowerCase();
-      return hostname === site || hostname.endsWith(`.${site}`);
-    } catch (error) {
-      return false;
-    }
-  });
-  const matchesGroup = !parsed.groups.length || parsed.groups.some((group) => group.every((term) => text.includes(term)));
-  const excludesTerm = parsed.excluded.some((term) => text.includes(term));
-  return matchesSite && matchesGroup && !excludesTerm;
-}
-
 function formatDate(value) {
   const date = new Date(value);
   if (Number.isNaN(date.getTime())) return value || 'updated recently';
@@ -128,6 +68,17 @@ function formatDate(value) {
   return `${date.getFullYear()} ${months[date.getMonth()]} ${date.getDate()} - ${hours}:${minutes}`;
 }
 
+async function loadConfig() {
+  const response = await fetch('/api/public/config');
+  if (!response.ok) throw new Error('Could not load config');
+  const config = await response.json();
+  pageSizes = config.feed_page_sizes?.length ? config.feed_page_sizes : pageSizes;
+  const pref = readPreferences();
+  const selected = pageSizes.includes(pref.pageSize) ? pref.pageSize : pageSizes[0];
+  pageSizeSelect.replaceChildren(...pageSizes.map((size) => new Option(`${size} per page`, String(size))));
+  pageSizeSelect.value = String(selected);
+}
+
 async function loadUsers() {
   const response = await fetch('/api/public/users');
   if (!response.ok) throw new Error('Could not load users');
@@ -237,6 +188,43 @@ function renderFeed(items, showIdentity = true) {
   });
 }
 
+function renderPagination(page, totalPages) {
+  paginationEl.innerHTML = '';
+  if (totalPages <= 1) return;
+
+  const goToPage = (target) => {
+    currentPage = Math.min(Math.max(target, 1), totalPages);
+    loadFeed();
+  };
+
+  const makeButton = (label, target, options = {}) => {
+    const button = document.createElement('button');
+    button.type = 'button';
+    button.textContent = label;
+    if (options.active) button.classList.add('active');
+    if (options.disabled) button.disabled = true;
+    button.addEventListener('click', () => goToPage(target));
+    return button;
+  };
+
+  paginationEl.appendChild(makeButton('Previous', page - 1, {disabled: page <= 1}));
+
+  const pageNumbers = new Set([1, totalPages, page, page - 1, page + 1]);
+  let previous = null;
+  [...pageNumbers].filter((num) => num >= 1 && num <= totalPages).sort((a, b) => a - b).forEach((num) => {
+    if (previous !== null && num - previous > 1) {
+      const ellipsis = document.createElement('span');
+      ellipsis.className = 'pagination-ellipsis';
+      ellipsis.textContent = '…';
+      paginationEl.appendChild(ellipsis);
+    }
+    paginationEl.appendChild(makeButton(String(num), num, {active: num === page}));
+    previous = num;
+  });
+
+  paginationEl.appendChild(makeButton('Next', page + 1, {disabled: page >= totalPages}));
+}
+
 async function postToMastodon(item, button) {
   button.disabled = true;
   const response = await fetch(`/api/links/${encodeURIComponent(item.id)}/mastodon`, {
@@ -321,35 +309,30 @@ function showEditForm(article, item) {
   article.appendChild(form);
 }
 
-async function loadFeed(selectedTag = null) {
+async function loadFeed() {
   const routeUser = document.body.dataset.userFilter;
-  const endpoint = routeUser
+  const pref = readPreferences();
+  const pageSize = pageSizes.includes(pref.pageSize) ? pref.pageSize : pageSizes[0];
+
+  const params = new URLSearchParams();
+  if (pref.tag) params.set('tag', pref.tag);
+  if (pref.search) params.set('search', pref.search);
+  params.set('sort', pref.sort || 'newest');
+  params.set('page', String(currentPage));
+  params.set('page_size', String(pageSize));
+  if (pref.user && !routeUser) params.set('username', pref.user);
+
+  const path = routeUser
     ? `/api/public/feed/${encodeURIComponent(routeUser)}`
     : '/api/public/feed';
-  const response = await fetch(endpoint, {
+  const response = await fetch(`${path}?${params.toString()}`, {
     headers: accessToken ? {Authorization: `Bearer ${accessToken}`} : {},
   });
   const data = await response.json();
 
-  let items = data || [];
-  const pref = readPreferences();
-  const activeTag = selectedTag ?? pref.tag;
-
-  if (pref.user && !routeUser) {
-    items = items.filter((item) => (item.user?.username || '').toLowerCase() === pref.user.toLowerCase());
-  }
-
-  if (activeTag) {
-    items = items.filter((item) => item.tags?.some((tag) => tag.toLowerCase() === activeTag.toLowerCase()));
-  }
-
-  items = items.filter((item) => matchesSearch(item, pref.search || ''));
-
-  if (pref.sort === 'oldest') {
-    items = [...items].reverse();
-  }
-
-  renderFeed(items, !routeUser);
+  currentPage = data.page || 1;
+  renderFeed(data.items || [], !routeUser);
+  renderPagination(currentPage, data.total_pages || 1);
 }
 
 function syncPreferences() {
@@ -362,6 +345,7 @@ function syncPreferences() {
   sortSelect.addEventListener('change', (event) => {
     const next = { ...readPreferences(), sort: event.target.value };
     writePreferences(next);
+    currentPage = 1;
     loadFeed();
   });
 
@@ -375,15 +359,29 @@ function syncPreferences() {
   tagFilter.addEventListener('change', (event) => {
     const next = { ...readPreferences(), tag: event.target.value };
     writePreferences(next);
-    loadFeed(event.target.value);
+    currentPage = 1;
+    loadFeed();
   });
 
+  let searchDebounce;
   searchInput.addEventListener('input', (event) => {
     const next = { ...readPreferences(), search: event.target.value };
     writePreferences(next);
+    currentPage = 1;
+    clearTimeout(searchDebounce);
+    searchDebounce = setTimeout(() => loadFeed(), 300);
+  });
+
+  pageSizeSelect.addEventListener('change', (event) => {
+    const next = { ...readPreferences(), pageSize: Number(event.target.value) };
+    writePreferences(next);
+    currentPage = 1;
     loadFeed();
   });
 }
 
 syncPreferences();
-Promise.all([loadUsers(), loadTags()]).then(() => loadFeed()).catch(() => loadFeed());
+loadConfig()
+  .then(() => Promise.all([loadUsers(), loadTags()]))
+  .then(() => loadFeed())
+  .catch(() => loadFeed());
diff --git a/frontend/static/style.css b/frontend/static/style.css
index bd255fc..10642d1 100644
--- a/frontend/static/style.css
+++ b/frontend/static/style.css
@@ -886,6 +886,53 @@ button:disabled {
   padding-bottom: 40px;
 }
 
+.pagination {
+  display: flex;
+  flex-wrap: wrap;
+  align-items: center;
+  justify-content: center;
+  gap: 4px;
+  max-width: 100%;
+  margin: 0 0 40px;
+  overflow-x: auto;
+}
+
+.pagination button {
+  min-width: 28px;
+  padding: 3px 7px;
+  background: transparent;
+  border: 1px solid transparent;
+  border-radius: 6px;
+  color: var(--muted);
+  font-size: 0.82rem;
+  line-height: 1.4;
+  cursor: pointer;
+}
+
+.pagination button:hover:not(:disabled) {
+  border-color: var(--border);
+  color: var(--text);
+}
+
+.pagination button:disabled {
+  opacity: 0.4;
+  cursor: default;
+}
+
+.pagination button.active {
+  background: var(--surface-1);
+  color: var(--text);
+  border-color: var(--border);
+  font-weight: 600;
+}
+
+.pagination .pagination-ellipsis {
+  padding: 3px 2px;
+  color: var(--muted);
+  font-size: 0.82rem;
+  opacity: 0.6;
+}
+
 .link-item {
   padding: 11px;
   background: var(--surface-0);
diff --git a/frontend/templates/feed.html b/frontend/templates/feed.html
index ba14223..74c4ef4 100644
--- a/frontend/templates/feed.html
+++ b/frontend/templates/feed.html
@@ -61,6 +61,10 @@
                 Search
                 <input id="search-input" type="search" placeholder="Search links" autocomplete="off" aria-label="Search links" />
               </label>
+              <label>
+                Per page
+                <select id="page-size-select"></select>
+              </label>
             </section>
           </div>
         </div>
@@ -84,12 +88,13 @@
       </section>
       {% endif %}
       <section id="feed" class="feed" aria-live="polite"></section>
+      <nav id="pagination" class="pagination" aria-label="Feed pages"></nav>
     </main>
     <footer class="site-footer">LinkLog Version {{ app_version }}. Copyright © 2026 Olaf Kolkman  · <a href="https://git.kolkman.org/olaf/Link-Log">git.kolkman.org/LinkLog</a></footer>
 
     <script src="/static/auth-header.js?v=3"></script>
     <script src="/static/logout.js?v=3"></script>
     <script src="/static/theme.js?v=1"></script>
-    <script src="/static/feed.js?v=13"></script>
+    <script src="/static/feed.js?v=14"></script>
   </body>
 </html>
diff --git a/frontend/version.json b/frontend/version.json
index 9c10763..085e3ba 100644
--- a/frontend/version.json
+++ b/frontend/version.json
@@ -1,3 +1,3 @@
 {
-  "version": "0.2.1
+  "version": "0.3.0"
 }
\ No newline at end of file