See your AI's changes work before your users do.
Say how a feature should evolve. Your own coding agent builds it. You try it as an admin, a new user and a locked-out account. You stay in the loop until it works.
Sign in with GitHub, Google or email. Works with Claude Code, Codex, Aider and any agent you mention on a GitHub pull request.
Features
How it evolved
- Email invites
- Bulk invites
- Next: Choose a role
What should change next
Admins choose a role when they invite someone, so new people don't all get full access
Before you merge
- Your repositoryConnect a GitHub repo or start from scratch
- Your coding agentAdaptl writes the brief, the agent writes the code
- No real customer dataEvery person in a preview is made up
- You stay in the loopNothing ships without your team's yes
From idea to merged change in three steps
The same loop for a one-line fix or a new feature, whether you work alone or with a team.
Describe it in your users' words
Write what should change the way a customer would notice it. Adaptl asks about the cases you missed, such as what a suspended account should see, then opens the issue, branch and draft pull request on GitHub.
Adaptl has your agent build it
Adaptl hands the brief to the agent you already use: Claude Code, Codex or Aider on your computer, or any agent you mention on the pull request. No new agent to trust, and your own AI key.
Try it as your customers, then merge
Open a preview full of made-up customers who feel real: an admin with years of invoices, someone who signed up today, an account locked after a failed payment. Pin what's wrong on the screen, the agent fixes it, you check again. That's the loop, and it runs until the change is right. Then your team merges.
1The line items overflow on a phone.
2The total ignores the VAT on the discount.
3"Send invoice" should ask for confirmation above €5,000.
Built for the part after the code is written
Coding agents are fast. Knowing whether what they built is right still takes a person. Adaptl gives that person what they need.
Previews that look like a business in use
An empty test app hides most problems. Adaptl fills every preview with made-up companies, people and history, written through your app's own models, so the rules in your code still hold. Switch between an admin, a brand-new user and a locked-out account in one click.
- Sanne de Vries AdminAcme BV · 4 years of invoices, 38 suppliers, 2 open approvals
- Tom Okafor New userBrightline Logistics · signed up today, nothing set up yet
- Lena Meyer Locked-out accountDefunct BV · last payment failed, read-only since 12 Oct
Your product's story, in plain English
Adaptl reads your merged pull requests and tells how the product evolved: what changed, why, and who it affects, grouped into chapters instead of a list of commits. Someone new can catch up in an afternoon.
- 7Anyone can start, and the team is set up properly31 Oct · 27 pull requests
- 6Adaptl goes online for teams30–31 Oct · 17 pull requests
- 4From idea to merged change, without leaving the app27–30 Oct · 18 pull requests
- 2Try a change before it merges, and get fixes back26–27 Oct · 8 pull requests
A living spec, written by your changes
Before a change merges, Adaptl lists what it changes for your users: what's new, what works differently and what goes away. Every line says what it's based on, so nothing is made up, and your team marks it right or puts it in their own words. Merge, and the lines you kept become part of what your product does today. Nobody writes a spec up front, and it never goes out of date.
- New: Admins choose a role for each person they inviteBased on: outcome · Invite screen
- Changes: New people start as Members, not adminsToday: everyone invited is an admin
- Goes away: The "make everyone admin" settingBased on: decision 1
What Acme does today · Team invites
- Admins invite people by email #12
- Admins invite a whole list at once #31
- Admins choose a role for each person #47
Helper agents that learn your project
Next to the agent that writes the code, Adaptl runs helper agents that watch what matters after: how fast your pages load, the errors your users hit, what your infrastructure costs, and how each change holds up in review. Each starts from a shared expert version and becomes yours: it remembers what it learns about your project, and you can adjust its rules for your team.
- Speed and memory WorkingChecks load time, click delay and memory on every change · 3 lessons kept for this project
- Code review ComingYour team's rules on top of the shared checklist
- Errors ComingTurns what your users hit into a change, with the evidence attached
- Infrastructure cost ComingFlags a change that would make hosting more expensive
Feedback the agent can act on
Each pin is tied to the spot on the screen and the customer you were signed in as, so the agent fixes the right thing.
Share a change with a client
Send a review link that opens one change only. It expires, and you can withdraw it at any time.
Know what your AI costs
See what your team spends on AI, on a key you control.
Checks before merge Early
Signed in as each made-up customer, Adaptl tries to reach other customers' data and flags any leak it finds. Running on Adaptl's own previews today.
For product teams: try the change, skip the diff
When AI writes most of the code, reading every line stops working. Adaptl gives each person on the team what they need to say yes.
"I want to try the change, not read about it."
Product managers describe the outcome, try it as different customers and steer with pins, without opening an editor.
"Show me what it touches, not 400 lines."
Tech leads see which screens and tables changed, whether migrations are safe and which journeys pass, and open the code when they want to.
"I built it with AI. Now it has to keep working."
Founders and agencies check every change as the product and its customers grow, and show clients a change before it ships.
Made for apps that already have years behind them
Adaptl works inside the product you already run, with its tables, roles and rules, not around it. We test it on Saleor and Cal.com, two large open-source products.
Where Adaptl fits
App generators start products and coding agents write code. Adaptl works with both and covers what comes next: checking each change before it reaches your users.
| App generators Lovable, v0 | Coding agents Claude Code, Codex, Cursor | Preview deployments Vercel, Render | Adaptl | |
|---|---|---|---|---|
| Main job | A first version from a prompt | Write the code | A link to each branch | Check every change before it ships |
| Data in the preview | Empty or mock data | Whatever is on your machine | Usually empty | Realistic made-up customers |
| How you review | Look at the result | Read the diff | Click through the link | Try it as each customer, pin feedback |
| Works with your agent | Uses its own | It is the agent | Not involved | Yes: sends it the brief and the pins |
Questions
Something else? Use the chat in the corner, or leave your email below.
What is Adaptl?
Adaptl keeps you in the loop while AI builds your product. You describe a change, Adaptl has your own coding agent build it, and you try it in a preview full of realistic, made-up customers. You pin what's wrong, the agent fixes it, and your team merges when it works.
Which AI coding agents work with Adaptl?
Claude Code, Codex CLI and Aider running on your own computer, or any agent you can mention on a GitHub pull request. Adaptl writes the brief and checks the result; the agent writes the code. You keep the tools and AI keys you already use.
Do I have to write a spec first?
No. Describe the change in your users' words. Before it merges, Adaptl lists what it changes for users, each line based on the change itself, and your team corrects anything that's off. Merged changes build an up-to-date description of what your product does, so the spec writes itself as you ship.
Do I need to read code?
No. You describe changes in plain words, try them in the running app and pin feedback on the screen. Engineers can open the pull request and the code whenever they want.
Does Adaptl use my production data?
No. Adaptl reads your code and database structure, never your customers' records. Every person and record in a preview is made up, written through your app's own models so the rules in your code still hold.
Can I start a new app with Adaptl?
Yes. Bring an existing GitHub repository, or start a project from scratch and let your coding agent build the first version. Either way, every later change goes through the same loop: preview, pin, fix, merge.
What does it cost, and where does it run?
Signing up is free. Adaptl runs online at inloop.adaptl.ai, hosted in the EU, or on your own computer. You use your own AI key, so your AI provider bills you directly for the model.
Who merges the change?
Your team, always. Adaptl prepares a normal pull request on GitHub. Nothing merges until a person on your team approves it.
Try your next change as your customers will.
Connect a repository, or explore a shared project in the playground first. Setting up takes about a minute.
Want a hand getting started?
We can set up Adaptl and your coding agent on your codebase with you and ship the first changes together. Leave your email and we'll get in touch.
We build Adaptl with AI coding agents and check every change to it in Adaptl itself.