
Short answer
Most extension idea lists stop at the name. The part that decides whether you finish is the manifest: which permissions the idea needs, what warning Chrome shows the user at install, and which store rule could get it rejected. Each idea below lists those, plus what to build first and what to watch out for. Platform rules come from Chrome's own developer docs, read in November 2026. If you would rather browse finished extensions before building one, our list of AI browser extensions for developers is a good tour of what already exists.
| Idea | Who it is for | Difficulty | Core features | Permissions |
|---|---|---|---|---|
| New tab dashboard | Anyone who opens 50 tabs a day | Beginner | To-do list, links, clock, quick notes | storage, plus the chrome_url_overrides manifest key |
| Tab session saver | Researchers, students, people who hoard tabs | Beginner | Save a window of tabs and groups, restore later | tabs, tabGroups, storage |
| Highlight and notes saver | Readers, writers, students | Beginner | Right-click to save a quote, search, export | activeTab, scripting, contextMenus, storage |
| Focus mode site blocker | Anyone who loses afternoons to feeds | Beginner to Intermediate | Block list, timed sessions, end notification | declarativeNetRequest, alarms, storage, notifications |
| Test data form filler | Developers and QA testers | Beginner | Fill forms with fake but valid data, saved profiles | activeTab, scripting, storage |
| Job application tracker | Job seekers | Intermediate | Save a job post, track status, reminders | sidePanel, contextMenus, activeTab, scripting, storage |
| Accessibility quick check | Front-end developers, content editors | Intermediate | Flag missing alt text, skipped headings, unlabeled fields | activeTab, scripting |
| Page change watcher | Shoppers, recruiters, anyone waiting on a page | Intermediate | Check a page on a schedule, notify on change | alarms, notifications, storage, optional host permissions per site |
| Side panel page summarizer | Heavy readers | Intermediate | Summarize the open page with on-device AI | sidePanel, activeTab, scripting |
| Select-to-translate helper | People reading in a second language | Intermediate | Detect language, translate selection on device | contextMenus, activeTab, scripting |
| Pull request review checklist | Developers and reviewers on GitHub | Intermediate | Team checklist on PR pages, saved per repo | Content script on https://github.com/*, storage |
| Private writing helper | Writers who don't want text leaving the laptop | Advanced | Rewrite or shorten selected text with on-device AI | sidePanel, contextMenus, activeTab, scripting, storage |
Chrome extension tutorials from a few years ago will lead you astray. Chrome's Manifest V2 timeline says Manifest V2 was disabled for all users with Chrome 138, and the remaining Manifest V2 extensions were removed from the Chrome Web Store on August 31, 2026. So every idea here assumes Manifest V3. Five rules shape the scope:
activeTab is your friend: it grants temporary access to the current tab when the user clicks your extension and shows no install warning.chrome.storage and use chrome.alarms for timers, not variables and setTimeout.Two numbers worth keeping in mind: chrome.storage.local holds 10 MB (more with the unlimitedStorage permission) and chrome.storage.sync about 100 KB, with 8 KB per item. Settings go in sync; collected data goes in local.
Replace Chrome's new tab page with your own: today's three tasks, pinned links and a scratchpad. Build first: a single HTML page registered under chrome_url_overrides, with tasks saved to chrome.storage.sync so they follow you between computers. Data: none beyond what you type. Watch out: an extension can override only one Chrome page, the New Tab override doesn't apply in incognito windows, and the store policy says to replace the new tab only through this documented override and by no other method.
Save every tab in a window, with its tab groups, as a named session and restore it next week. Build first: save and restore one window. Data: tab URLs, titles and group names, stored locally. Watch out: reading tab URLs needs the tabs permission, which shows the warning "Read your browsing history", and tabGroups shows "View and manage your tab groups". Say clearly in the listing that sessions never leave the device.
Select text, right-click, save. The popup lists every quote with its source page and exports to Markdown. Build first: the context menu item and the saved list. Data: the selected text, page title and URL. Watch out for pages where injection is blocked: activeTab doesn't grant access to restricted pages such as chrome:// pages, so handle that case with a friendly message. There is a ready prompt for this one below.
Block your worst sites for a timed session, then lift the block automatically. Build first: a static block list and a start button. declarativeNetRequest blocks or modifies requests through rules you register, without intercepting them or viewing their content, though it still shows the warning "Block content on any page". Data: the list of domains and the session end time. Watch out for timers: Chrome caps alarms at once per second and may delay them, so store the end time rather than counting down in memory.
One click fills sign-up and checkout forms with fake but valid data: names, emails on a domain you control, test card numbers your payment provider publishes. Build first: detect inputs by type and label on the active tab and fill them. Data: a few saved profiles. Watch out: keep it a developer tool. Never store real personal or payment data in it, and match fields by label as well as name, since many forms use random IDs.
Right-click a job post to save the title, company and link into a side panel board with columns such as Saved, Applied, Interview and Offer. Build first: saving from the context menu and a status dropdown. The side panel API needs Chrome 114 or later and the sidePanel permission. Data: what the user saves, plus their own notes. Watch out for scraping every job site's layout. Save what the user selects instead, and let them edit it.
Click the icon and the page lights up: images without alt text, heading levels that skip, form fields without labels, buttons without names. Build first: three checks and an overlay that outlines each problem. Data: the page's own DOM, read on demand through activeTab. Watch out for overselling. A quick check finds obvious gaps; it doesn't certify a page as accessible, and the extension should say so.
Watch a product page, a job board or a docs page and get a notification when the part you care about changes. Build first: one page, one CSS selector, a check every hour with chrome.alarms. Request each site's host access at runtime through optional_host_permissions instead of asking for every site at install. Watch out for the site's terms and for polling too often. Hourly is plenty for most pages.
Open the side panel, press Summarize, read the key points of a long article without leaving it. Chrome's built-in AI APIs list the Summarizer API as available to extensions from Chrome 138, running on the device. Build first: extract the article text and show key points. Data: the open page only. Watch out for hardware: the docs require Windows 10 or 11, macOS 13 or later, Linux or a Chromebook Plus, and at least 22 GB of free storage, and the APIs aren't on Android or iOS yet. Check availability before you show the button. A ready prompt is below.
Select a sentence, right-click, read it in your language. Pair the Language Detector API (to find the source language) with the Translator API; Chrome lists both for extensions from Chrome 138. Build first: one target language and a small popup with the result. Data: the selected text, which never leaves the device. Watch out for the first run: the translation model downloads the first time the API is used, so call availability() and show download progress instead of a frozen button.
Inject your team's review checklist into GitHub pull request pages: tests added, migrations reviewed, docs updated, screenshots attached. Build first: a content script that matches only https://github.com/* and shows a collapsible checklist saved per repo. Data: the checklist items and ticked state. Watch out: GitHub's page structure changes, so anchor the panel loosely and fail quietly. Keep the host match to github.com; asking for all sites would slow review and scare users.
Select text in any text box and ask for a shorter, clearer or more formal version, generated on the device. Chrome lists the Prompt API for extensions from Chrome 138; the Writer and Rewriter APIs are still developer trials, so don't build on them for a public release. Build first: rewrite a selection into the side panel, with a Copy button. Data: the selected text and the user's chosen tone. Watch out for writing straight into page fields: show the result and let the user paste it. Same hardware limits as the summarizer.
Want ideas that call a hosted model instead of the browser's? Our AI project ideas cover retrieval, vision and speech projects, and the backend half of any of them can sit behind an extension.
Want to try one of these? Start on Greta.sh — no setup, no boilerplate — just a prompt.Paste these into an AI coding tool or chat model. Each one names the permissions up front, which stops the model from requesting <all_urls> out of habit. Replace everything in [square brackets], then load the result unpacked from chrome://extensions with Developer mode on.
A good first build. Paste it into an AI coding tool and ask for the files one at a time if the tool tends to rush.
Build a Chrome extension using Manifest V3 called [Highlight Keeper]. Permissions: only "activeTab", "scripting", "contextMenus" and "storage". No host permissions. Behavior: when I select text on a page and choose "Save highlight" from the right-click menu, save the text, the page title, the URL and the date to chrome.storage.local. Popup: list saved highlights grouped by page, newest first, with a search box, a delete button per item and an "Export as Markdown" button that downloads a .md file. Do not send any data off my machine. Do not load any script from a URL; bundle everything. Give me manifest.json, the service worker, the popup HTML, CSS and JS, and a short README explaining how to load it unpacked from chrome://extensions.
Uses Chrome's built-in Summarizer API, so check that your machine meets the hardware requirements before you start.
Build a Manifest V3 Chrome extension that summarizes the current page in the side panel. Permissions: "sidePanel", "activeTab" and "scripting". Clicking the toolbar icon opens the side panel (use sidePanel.setPanelBehavior with openPanelOnActionClick). When I press "Summarize", inject a script into the active tab that returns the main article text (skip nav, footer and ads), then pass it to Chrome's built-in Summarizer API. Let me pick the summary type [key-points / tldr] and length [short / medium] before running it. First check whether the Summarizer API is available. If it is not, show a clear message that explains why instead of failing silently. If the model needs to download, show progress. Keep everything on the device. No network calls, no remote code.
Teaches declarativeNetRequest and alarms, the two APIs most Manifest V2 tutorials get wrong.
Build a Manifest V3 Chrome extension for focus sessions. Permissions: "declarativeNetRequest", "storage", "alarms" and "notifications". Options page: I enter a list of sites to block, for example [youtube.com, x.com, reddit.com], and a session length in minutes. Popup: a Start button. When a session starts, add dynamic declarativeNetRequest rules that redirect those sites to a bundled blocked.html page showing the time left. Set a chrome.alarms alarm for the end of the session; when it fires, remove the rules and show a notification. Store the session end time in chrome.storage so it survives the service worker shutting down. Do not use setTimeout for the session timer. Add a "Stop early" button that asks me to type "stop" before it ends the session.
Start with your own browser. The best first extension fixes something you do ten times a day, because you'll test it ten times a day. Then run it through three filters:
activeTab or a single host. Chrome's review process page says broad host permissions can make review take longer, and the install warning costs you users.chrome.storage need no server, no accounts and a much shorter privacy policy. Add a backend only when sync across people, not devices, is the point.Building for a weekend event? Our hackathon project ideas are scoped for 24 to 48 hours, and the focus blocker or highlight saver from this list fit that window too. If you want something with a full web app behind it, browse our web app ideas or the vibe coding projects list.
What is a good first Chrome extension to build?
A new tab dashboard or a highlight saver. Both need only a handful of permissions, run without a server, and teach you the manifest, the popup or page, and chrome.storage in a weekend. Save the tab session saver for second, because it needs the tabs permission and a clear privacy story.
Do new Chrome extensions have to use Manifest V3?
Yes. Chrome's own timeline says Manifest V2 extensions were disabled for all users on all channels with Chrome 138, and that all remaining Manifest V2 extensions were removed from the Chrome Web Store on August 31, 2026. Start every new project with manifest_version 3.
How much does it cost to publish on the Chrome Web Store?
Chrome's registration page says you pay a one-time registration fee when you first open the developer dashboard and agree to the developer agreement. The page does not state the amount, so check the dashboard before you budget.
Can a Chrome extension call an AI model?
Yes, two ways. It can call a hosted model through your own backend, which keeps the API key off the user's machine. Or it can use Chrome's built-in AI APIs: Chrome's docs list the Summarizer, Translator, Language Detector and Prompt APIs as available to extensions from Chrome 138, on Windows 10 or 11, macOS 13 or later, Linux and Chromebook Plus devices with at least 22 GB of free storage.
Which permissions slow down Chrome Web Store review?
Chrome's review process page says reviews may take longer for extensions that request broad host permissions such as <all_urls> or https://*/*, sensitive execution permissions, or a lot of hard-to-review code. Most reviews finish within a few days, but some take up to a few weeks.
Can I load code from my server into a Manifest V3 extension?
No. Chrome's docs say Manifest V3 extensions must bundle all the code they run, and the store calls anything else remotely hosted code. You can still fetch data such as JSON from your server. The Chrome Web Store also bans obfuscated code, though minification is allowed.
© 2026 Questera
Follow in Google