Five Launches From October 30: Travel Software, Agent Readiness, and a Pocket Money App

Yesterday’s batch didn’t cluster around a single theme, which is what makes it worth a look. Among the new developer tools and consumer apps that went live on October 30, 2026, we got self-hosted travel booking infrastructure, a scanner that checks whether AI agents can actually use your website, a family finance app built around three jars, a managed outbound agency, and a Mac client for ChatGPT Dots. Community voting is still open on all five, so no rankings have settled yet. Here’s what each one actually does, and where I’d push back.

PHPTRAVELS

If you run a travel agency or a small OTA and you’ve priced out booking platforms, you know the licensing math gets ugly fast. PHPTRAVELS takes the opposite approach: one-time licence from $2,499, full source code, self-hosted on your own PHP and MySQL stack. You get a white-label B2C site, a B2B agent portal, and engines for flights, hotels, tours and cars, with live GDS, NDC and bedbank supplier connections wired in.

The appeal is obvious for anyone who has been quoted per-booking fees or watched a SaaS vendor raise prices. You own the thing. The tradeoff is equally obvious: self-hosted means you’re the one patching it, and supplier integrations are where these projects usually get complicated. A demo and documentation are both public, so you can judge the code quality before committing. Worth a serious look if you have an in-house PHP developer; less appealing if you don’t.

Dedica

Most sites were built for humans and quietly fail for software. Dedica tests your site the way an AI agent would use it, then hands your team a fix for every point of failure. Enter a domain, get 125 checks across discovery, access, usability and payments, scored out of 100. The free scan takes under a minute with no signup, and the detailed fix kit runs $30 one-time per site.

The failure modes it names are specific and recognizable: a bot check returning 403 on your signup page, pricing that only renders after JavaScript runs, no OpenAPI spec published at the usual paths, no MCP server. The sample report shows a site scoring 71 with a D grade and 42 fixable issues, each ranked by how much fixing it would move the score, with an estimated gain attached. Fix files are named to map directly to issues and written so you can paste them into Claude Code, Cursor or Copilot.

The 6%-of-sites-are-agent-ready figure is unverifiable marketing, and the 1.9× recommendation stat should be treated the same way. But the underlying diagnosis is sound: when an agent gives up on your signup flow, nothing logs an error. That invisibility is the real problem, and $30 to audit for it is cheap.

TONKO

A pocket money app for families with kids aged 6 to 14, built around a simple mechanic. Parents set the allowance and the jobs. Kids earn by completing work that a parent approves with one tap, then split every euro between three jars: Spend, Save and Grow. Content adjusts across three age bands. No ads, nothing sold, and no real money held — TONKO is a ledger over pocket money you already hand out.

The design choices are thoughtful. Approval gates mean nothing moves without a parent, a shared phone hands off to a child’s space and closes with a fingerprint or PIN, and kids use a nickname and avatar rather than an email and password. The Money Time Machine lets a child drag sliders and watch €5 a week compound into something concrete, which is a better teacher than a worksheet.

Two caveats. It’s Android-only for now with iOS planned, which cuts out a lot of families. And it speaks English and Croatian, following the phone’s language. Free with one child, so trying it costs nothing.

throxy

A done-for-you outbound service for B2B companies selling into manufacturing, logistics, construction, education and healthcare. throxy builds prospect lists with its own scraping and trigger-detection tooling, then puts human SDRs on the phone and in the inbox. It calls itself Y Combinator’s first AI-native sales agency, which is the kind of claim to file away and judge on delivery.

The mechanics are the interesting part. Lead research pulls from LinkedIn, prospect websites and Glassdoor looking for timing triggers like a fleet expansion. The dialer, Maestro, routes calls to an SDR speaking the lead’s language. Qualification runs against four criteria, and every meeting goes through an approval step — they state plainly that nothing gets booked without your sign-off. One customer reports 12 meetings in 20 days in December, which is genuinely hard.

Pricing is the blank spot. No tiers, no list prices, just a strategy call booking form. The stated model is paying for meetings you accept that actually happen, which is a fair structure in theory but leaves the actual number invisible until you’re on the call. Expect a real sales conversation, not a self-serve signup.

Dotverse

A Mac app for people who already have a Dot in ChatGPT and want to organize it. Instead of a separate AI service or an API key, Dotverse connects through a sign-in window and lets you build a crew of custom Dots, each with its own role, personality and conversation view. The examples given are a candid editor for writing, a pragmatic planner for planning, and a patient tutor for learning.

Custom Dots share the underlying Dot’s memory and capabilities, so you don’t rebuild context for each role. Requests run one at a time. Encrypted copies of conversations and drafts stay on your Mac, though messages still go to the connected Dot service, and clearing local history doesn’t clear Dot’s own memory.

It’s version 0.2.0, requires Apple Silicon and macOS 14+, and no pricing is listed anywhere on the site. Treat this as an early experiment rather than a finished product — the concept of task-specific personalities sharing one memory is genuinely useful, but the app is thin right now and depends entirely on a service you already pay for.

Yesterday's Top Launches: 5 Tools from October 30, 2026