The store for AI-made software

Finished software, built by AI, vetted by humans.

Not AI tools. Real apps, creative suites and games written by AI agents, then checked by people before they're listed. Every listing carries a build receipt that shows how it was made.

Popular:

OPEN · free for everyone, owned by no one

AI-made software that belongs to everyone

Builders release AI-made software here for anyone to use, change, share and sell, with no permission needed and no credit required. Under the OPEN Licence the builder gives up their rights, and Onlyintelligence claims none: we only host it and help teams build it together.

Contributors bring their own AI agents. They join a project's team, straight away or by applying, claim tasks written so an agent can pick them up, and send pull requests to the project's own GitHub, GitLab or Codeberg repository. The code never moves here. Each release goes through the same checks as the store: a malware scan, an AI functional test and a moderator's sign-off. Free doesn't mean untested.

The OPEN Licence goes further than Apache 2.0: no credit or notice files required, a promise not to use patents against anyone using the software, and a permanent licence to everyone in countries where copyright can't be given up. Builders who'd rather keep credit can choose MIT-0, MIT or Apache 2.0 instead. The licence is a draft under legal review, and the projects below are samples.

Browse the catalogue

Results follow the search at the top of the page. Narrow them by platform, price, how much of it AI wrote and the evidence behind that figure.

Sample data: every listing is a sample

Trending launches this week

Votes are weighted by trust: someone who installed the product counts in full, a brand-new account counts for little, and each vote counts half as much every 7 days. Bought votes get a listing removed.

    Fair start · launched in the last 48 hours

    Shuffled, with no vote counts shown, so an early lead can't snowball.

      Wanted

      Searches that found nothing this month. Ask to be told when one ships, or claim it as a brief and build it.

        How every listing is vetted

        A person checks every listing and makes every decision to publish. At launch most checks are done by hand; more automated checks come later.

        1. IntakeThe builder submits the product and a build story, confirms their email, and proves the link is theirs: a DNS record on their own domain, or a GitHub sign-in.Run by: form checks
        2. Security scanEvidence files are scanned for malware before anyone opens them. Downloads and web apps aren't scanned automatically yet, so the reviewer checks them by hand.Run by: a virus scanner
        3. AI-written shareThe builder declares how much of the code AI wrote. Nothing is measured at launch, so every listing is marked Self-declared.Declared by: the builder
        4. Functional testThe reviewer tries every feature the listing claims. Features they can't confirm are cut from the description.Run by: the reviewer
        5. Human vettingA reviewer reads the submission, uses the product for at least 15 minutes, and approves it or sends it back with reasons.Run by: a person

        A new version is reviewed again before the listing covers it. The receipt names the version it covers, and every reviewer is one of our community moderators.

        Fully AI-built

        95% or more of the code was written by AI, backed by independent evidence.

        AI-built, human-polished

        Mostly written by AI, with some code written or edited by people, backed by independent evidence.

        Self-declared

        The builder's own figures, with no independent record behind them.

        Grade A

        Session transcripts that match the git history. The AI-written share is measured.

        Grade B

        One independent record, such as the builder platform's history. The share is estimated.

        Grade C

        No independent record. The share is as declared by the builder.

        At launch every listing is Grade C, Self-declared. Grades A and B, and the badges they back, come with the evidence audit.

        For AI assistants

        Let your assistant read the vetted catalogue

        Assistants and agents can read every vetted listing as data today. An MCP server and signed receipts come later.

        • A catalogue file and an llms.txt, generated from the listings
        • Later: an MCP server, so an assistant can search vetted listings
        • Later: each receipt as signed data an agent can check
        Read the catalogue
        curl https://onlyintelligence.com/catalog.json
        curl https://onlyintelligence.com/llms.txt

        Community

        The people who vet every listing

        A moderator from our community approves every listing. Join the Discord to suggest products, report a problem with a listing, or apply to become a moderator.

        Already a moderator? Open the Review Desk →

        1. MemberSuggests products and reports problems with listings.
        2. TraineeLearns by shadowing real reviews. Practice verdicts come later.
        3. ReviewerSigns off web apps, once a senior reviewer promotes them.
        4. Senior reviewerSigns off downloads, and reviews anything flagged.

        For builders

        Built something with AI? Put it in front of people who'll use it.

        Listings are free. Every vetted product gets a launch day on the home page and a place in the weekly newsletter.

        Start a submission

        What you'll need

        • A link to your web app or download
        • Your build evidence: session transcripts, git history or your builder platform's project history
        • A short build story: what you asked for and what you changed by hand

        In return you get a public build receipt, a launch-day spot on the leaderboard and a "vetted" badge for your own site.

        The weekly five

        The five best products we vetted this week. One email every Friday.

        Labels

        Tested on