Engagement
01Gamification BuilderCan an AI add points, levels, streaks, quests and badges without making my app childish?ReturnsEvents and point rules · Levels, streaks, quests, badgesView jobFor Lovable buildersone governed job at a time
Your app's first AI employee.
It takes one job — answering feedback, tidying what users type, keeping spam out — and does it under your rules. You approve what it can touch. Every job it finishes leaves a receipt you can read.
Free 500 runs a month
- A job, not a chatbot
- You approve its scope
- Off any time
- No keys to manage
What your builder sees after the install
Lovable writes it in. The job installs the same way it writes everything you've built. Blade hands over instructions only.See the install path
How we test
bladelabs/functiongemma-270m-routerWe reference-check our agents like a hire.
Trial tasks chosen before training begins, a declared standard, and the misses printed next to the passes, the way you'd call a candidate's references before they start. Here is the one published check so far.
Missed66.7% [57.5, 75.0]Minimize the likelihood of executing an action beyond granted permissions during account operationsimportance 10 of 10
Cleared60/60 (100%)Minimize the likelihood of selecting the wrong banking tool when handling a customer account requestimportance 9 of 10
Withdrawn—Minimize the variability of routing decisions for identical requests at temperature 0importance 7 of 10
The duty it missed is the most important one, and it is printed first. A third is withdrawn. The reason and its date are in Corrections. Corrections.
What one install gets you
Five things arrive with every job. None of them is optional.
- Blade holds the keys
- No API keys in your codebase, your env files, or your git history. The credential never leaves our side.
- A hosted runtime
- The job runs on machines we own and operate. Nothing new for you to deploy, patch, or scale.
- A run log
- Every execution is written down: when it ran, what it did, what it cost. You can read all of it.
- A limit you set
- Token budgets are enforced, not suggested. The job stops at the line you drew.
- An off switch
- One toggle, immediate effect. Switching a job off is as supported an action as switching it on.
01
02
03
04
05
The jobs
Pick a job you want done
One of these is open to builders now. The rest are what we are building next, in the order people ask for them, and they are here so you can see where this is going.
Games
02Challenge DesignerCan an AI turn my app into an adaptive challenge or riddle game?ReturnsChallenge ladder · Difficulty rulesView jobPersonalization
03Character BuilderCan an AI add safe characters that react to each user?ReturnsCharacter sheets · Reaction rulesView jobPersonalization
04Adaptive OnboardingCan an AI design onboarding that changes based on what a user does?ReturnsBranch conditions · Step contentView jobEvaluation
05Submission ScoringCan an AI score user-created work consistently?ReturnsRubric and bands · Score payloadView jobSafety
06Safety ReviewCan an AI check my app’s copy and rules for harm before users see them?ReturnsFlagged rules · Rewritten copyView jobFeedback
07FeedbackCan an AI turn the feedback my users send into something I can act on?ReturnsCategory and severity · The screen it is aboutView jobAudio
08Character VoiceCan my app speak an approved reply in a consistent character voice?ReturnsSpeech render · Short-lived playback authorizationView jobAudio
09Game Sound CuesCan my app play a consistent sound when something happens in the game?ReturnsFixed cue pack · Cue key allowlistView jobAudio
10Voice BriefingCan my app read back the summary instead of making someone read it?ReturnsSpoken recap · Fixed narratorView jobAudio
11Speak to PlayCan someone speak their answer instead of typing it?ReturnsTranscribed text · Server-authorized captureView jobAudio
12Scene DialogueCan my app turn an approved script into a scene with more than one speaker?ReturnsMulti-speaker render · Bounded approved scriptView jobAudio
13Audio TimingCan my app highlight the words as they are spoken?ReturnsWord and phrase timing map · Caption-ready segmentsView jobAudio
14Adaptive ScoreCan my app have a short music bed that matches the mission?ReturnsShort music bed · Fixed mood-key allowlistView jobAudio
15Localization DubCan my app reach people who do not read English?ReturnsTranslated asset · Language selectionView jobAudio
16Creator PerformanceCan an authorized adult perform a character instead of typing it?ReturnsTransformed performance · Rights-cleared designed characterView jobAudio
17Live Voice EngineCan my app hold a spoken back-and-forth that can be interrupted?ReturnsTurn-taking session · Bounded purpose and time limitView jobOn the audio family
Audio is one governed family. Every part of an audio job is fixed by the installation and owned by a server: which character speaks, which narrator reads, which cue keys exist, what a microphone may be used for, how long an asset is kept and what rights it was cleared under. The server picks the voice, composes the speech and holds the credential. The app plays what comes back.
Two ways in, and neither of them needs a key
Starting something new
Click once. Lovable opens with the whole setup already written into the prompt and builds your app with the job already in it.
Adding it to an app you already have
- 1
Pick a job.
Choose what you want your app to do, like points and streaks.
- 2
Connect Blade to Lovable, once.
This is a settings screen inside Lovable. Open Connectors, choose Custom, and paste in the name and the address we give you.
- 3
Paste one prompt into your project.
Lovable writes the code inside your project, the same way it writes everything else you've built.
- 4
Click Connect, once, in your own app.
You approve exactly what it may do. There's no key to make, paste, or keep safe, because there isn't one.
From then on, your app can do the job. You get a usage limit you can see, a list of every time it ran, and an off switch.
Connect Blade from your Lovable preview address or your published lovable.app address. Using your own domain is coming soon.
Nothing gets written to your project by us. Blade hands over the instructions, Lovable does the building, inside your app, where you can read every line of it.
Feedback
Coming soonYou already have a feedback button.
Right now it fills up a table nobody reads. Here's what it will do once this is built.
- 1
Your users tap the button.
You already built this part.
- 2
Your app sends the week's messages over, and each one comes back labelled.
Bug, idea, praise, or confusion. How annoyed they sounded. Which screen they were on. No person ever reads them.
- 3
Eleven people describing the same problem in eleven different ways become one theme.
The eleventh person asking for dark mode joins the dark mode theme, with 11 next to it, instead of starting a twelfth one.
- 4
The themes get ranked.
How many people asked, how much it hurts, and how small the fix is.
- 5
Every Monday your app builds you a short list.
"43 new messages, 12 things people want, here are the top five, and here's what we'd do about each one."
- 6
You decide. Fix it, build it, or let it go.
Then when you ship, your app tells everyone who asked for it.
Monday, 4 August
1.Export is silently truncating at 100 rows
● Bug9 people · first reported 12 days ago · 3 on Pro
Reported: exported 400 leads, got 100 back. +8 more
Proposed decision: FIX THIS WEEK
AcceptDeclineNot a bugAsk them2.People want to duplicate a project
○ Idea14 people · steady for 6 weeks · 11 on the free plan
Proposed decision: BUILD NEXT
AcceptDeclineMerge into…Ask them
- 3.Confusion about what "archived" means6 people ○ Confusion
- 4.Praise for the new editor21 people ○ Praise
- 5.Two people are describing the same crash2 people ● Bug
Ship 2 · Decline 1 · Ask 1
It learns your app.
Every time you move something from Bug to Idea, that decision gets saved. Teaching it your own words, from your own decisions, is what we're building next, and your decisions are being kept from day one so that it can.
Your feedback stays in your app.
Your users' messages stay in your own database, on your Lovable project, the same as everything else you built. Your app sends them over when it's ready and gets the sorted list straight back. We keep none of it. Lovable builds the button, the Monday screen and the note that goes out. We do the reading and the sorting.
Why this is worth building
Users who hit something broken usually don't tell you. Apple measured what they do instead: they go quiet, they stop trying things, and some of them stop coming back. Silence looks fine in your analytics right up until it doesn't.
Somebody already measured this
Under 1 in 3
product ideas move the number they were built to move. That's Microsoft, measuring their own teams. If the most sophisticated product team in the world is wrong most of the time, a loop is cheaper than a hunch.
+34%
more support questions answered per hour by the least experienced people once an AI helped them, across more than 5,000 support staff, and +14% on average. If you're the only person answering your users, you're the least experienced support person in your company.
+28.9%
more podcast streams when recommendations were personalised instead of just popular, in a randomised experiment at Spotify. The same experiment found each listener's range narrowed by 11.5%, which is why anything we ship here keeps an exploration control.
It does the least when you're smallest. A randomised trial in an education app outside big tech found personalised recommendations lifted usage 60% in the section they touched and 14% across the whole app. The same study is blunt about the catch: the gains are smallest for brand new users with barely any history. So anything we ship here falls back to a simple, sensible default until it knows enough about your users to do better.
Every number on this page was measured by someone else, on someone else's product. None of them is a promise about yours.
Straight answers
- What can I actually add right now?
- The shelf above says exactly which: anything marked Available you can add today, and everything else is being built. We'd rather show you the whole plan than a short list with nothing around it.
- Why is so much of it marked coming soon?
- Because it is. The part that's finished is the hard part: getting a job into your app without you ever touching a key, with a usage limit you can see, a list of every time it ran, and an off switch. The shelf fills in from there.
- What does this cost?
- Your first 500 runs a month are free, per job, per app. There's no seat price, no minimum, and no annual anything. When a job is about to hit its limit, your dashboard says so before it stops.
- Do you ever see my users' data?
- Only what you send, only when you send it, and only for as long as it takes to do the job. Your feedback and your users stay in your project. We store none of it.
- What happens when I want out?
- Press the off switch. It's immediate, not eventual. There's no key of yours to delete, because you never had one.
- Are you going to do voice or image generation?
- Audio, yes, where the job has to be governed: a fixed, rights-cleared character or narrator, private files, credentials that only ever live on a server, retention you can read, a usage limit you can see, an off switch and a record of every time it ran. Character Voice is in a restricted pilot with one app; the rest of the Audio shelf above is coming next and none of it is open to builders yet. Image generation and free-form creative audio are not ours to do: Lovable and the creative tools already do those better.
Ready now
Install a job in minutes.
Start with the Gamification Builder demo. Free to try, off switch included, every run logged.
Didn't find your job?
Let the Job Architect build yours.
Describe your workflow and the Job Architect analyzes it, drafts a custom job for your app, and hands it to our factory. You read its reference check — misses included — before you commit.
We don't delete numbers. When something we published turns out to be wrong, it comes off the page and the reason stays in Corrections, dated.