Clearly state what your extension does and get a valid manifest.json, the exact list of permissions, and the literal notice that users will see when they install it. Free, no account needed, and no AI.
You describe capabilities; the app deduces the manifest. No extra permissions, no surprise installation notices.
The Google Chrome Manifest V3 builder starts from what your extension needs to do, not from the API documentation. You select from 24 capabilities written in plain language —read the active tab, inject a script into a page, save data, add a context menu, open a side panel, set a keyboard shortcut, declarative network rules, alarms, clipboard, downloads, notifications, omnibox, OAuth— and the app deduces the minimum set of permissions from a catalog of 34, plus the host permissions for the specific domains you specify.
For each permission that goes into the manifest, it shows you two things: its risk level and, when Chrome shows a notice during installation, the literal text that the user will read. Marking `tabs` means 'Read your browsing history'; `history`, 'Read and change your browsing history on all signed-in devices'; a host permission, 'Read and change your data on' followed by your domains. Of the 34 permissions in the catalog, 16 are marked as high risk and 12 do not generate any notice, so you can quickly see which ones cost you installations and which are free in terms of trust.
From this, the complete manifest.json is generated with indentation: manifest_version 3, name, version, description, icons in 128/48/32/16, action, the service worker, and only the keys corresponding to what you've marked —content_scripts, options_ui, side_panel, commands, declarative_net_request with its ruleset, web_accessible_resources, omnibox, oauth2, minimum_chrome_version, host_permissions, and optional_host_permissions. Along with the JSON comes the project file tree and the initial skeleton for each: service-worker.js, content.js and content.css, popup.html/js/css, options.html/js, sidepanel.html/js, offscreen.html/js, rules.json, and a README with submission steps.
And before giving it to you, it validates against its own schema and the verified limits of the store: that the manifest version is 3, that the name does not exceed 75 characters and the short description 132 —counted by actual glyphs, not bytes—, that the version number consists of one to four numbers separated by dots, that the icons are declared. It also warns you about what slows down a review: requesting access to all sites instead of specific domains, declaring `tabs` and `activeTab` at the same time when only the latter would suffice, accumulating high-risk permissions, and the three requirements waiting for you on the dashboard —a justification for each permission, including host permissions, the privacy policy URL, and the remote code declaration.
The entire process is deterministic: no AI model intervenes, no third-party API is consulted, and no account is required. As a guest, you can generate up to 10 manifests in one session; a free account is only needed to save them with a name and download the project as a ZIP.
Use it free without an account. If you want to save your result, you can sign up afterwards.
Save / Sign upMinimum permissions by design
you declare capabilities, not permissions. The app translates the 24 capabilities into the essential permissions and does not add any 'just in case' —one of the checks literally tells you how many capabilities you have declared and that all permissions come from that, and none more.
Installation notice before publishing
for each permission that generates a notice, you see the literal text in English that Chrome will display upon installation, with the domains already substituted. It's the difference between deciding with data and discovering it when your submission is already under review.
Risk by permission, with the costly ones highlighted
the catalog classifies its 34 permissions into high, medium, and low risk —16 of them high— and the validator summarizes how many high-risk ones you have and which, because each will need a very specific justification in the dashboard.
Specific domains instead of `<all_urls>`
if what you mark requires host permissions, it asks for the exact domains and warns you as soon as it detects a pattern open to all sites, which is the number one reason for slow review or rejection. It also supports requesting them as optional.
Detects classic redundancy
if you request `tabs` and `activeTab` at the same time, it marks it and explains that removing `tabs` eliminates the notice 'Read your browsing history' when you only need the tab that the person has in front of them when they click your icon.
Limits counted by actual glyphs
name up to 75 characters, short description up to 132. They count as real characters, so an accent or an emoji counts as one, not two or four. It tells you how many you have out of how many, not just if you pass or not.
Manifest and skeleton, not just the JSON
in addition to manifest.json, it shows the project's file tree and the initial content of each one —service worker, content script, popup, options page, sidebar, offscreen document, network ruleset, and a README with submission steps— ready to open in your editor.
Deterministic, free, and no account needed
zero calls to AI and zero third-party APIs, so the same manifest is generated the same today and tomorrow. Up to 10 manifests per session as a guest; the free account is only for saving and downloading the ZIP.
Select what your extension does and view the manifest, permissions, and installation notices right there. No account, no AI, and no cost; create your free account only if you want to save it and download the project in ZIP.
In the features that use AI, responses are generated by artificial intelligence and are not a substitute for professional advice.