Limits and FAQ

Limits today

Limit	Value
Sign-in code	6 digits, valid one hour. Asking again within 10 minutes sends the same code.
Pending project	At most 100 frames (prototype and live), 400 versions and 10 MB of captures until it is claimed. All pending projects share 200 MB of captures.
Claim link and temporary key	7 days. Your agent can renew them (renew_claim_link) before they expire; once expired, the key stops working and the agent must create a new project.
New pending projects	50 per hour across the platform.
HTML of one version	102400 bytes (100 KB).
Request body	1048576 bytes (1 MB).
Capture image	2 MB, JPEG, PNG or WebP.
Web app	Designed for desktop screens.
Agents	Built and tested with Claude Code. Any MCP client or HTTP caller can use the tools with a key.

Who can sign in?

Anyone with an email address. The app sends a 6-digit code by email; there is no password. The first time, the app asks you to create an organization.

What if the sign-in email does not arrive?

The code usually arrives within a minute: look in your spam folder too. Send the code again sends the same code for 10 minutes, and a code stays valid for one hour. Codes asked for too often, from one network or for one address, are refused for a while. If nothing arrives, write to hello@logics-studio.com.

Why does my agent not add the MCP server itself?

Adding an MCP server changes your Claude Code configuration, and Claude Code only loads a new server at its next start. Your agent does not need it: it uses the HTTP API, with the same tools and the same project, for the whole session. At the end, it gives you a command to run in your own terminal if you want the MCP server in later sessions. After you run it, restart Claude Code with claude --continue, which keeps the conversation.

Claude Code stopped because an action was blocked. What now?

In auto mode, Claude Code can block actions that go beyond your request or reach a destination it does not know, such as adding an MCP server or sending screenshots your message did not mention. Say explicitly what to send and where (“send the screenshots to app.agentappbuilder.com”) and run the MCP command yourself if you want it. Your message cannot clear every block: if sending the screenshots is still blocked, describe app.agentappbuilder.com as a trusted destination in your user settings. Getting started has the details.

Does my agent capture my app without asking?

Right away only when your message names both the Playwright install and app.agentappbuilder.com, like the confirmation sentence below. Otherwise it first asks you, in one message: the screens it will capture, where the screenshots go, that your app must show demo data only, and the account it will sign in with. It ends with a sentence for you to send back, “Install Playwright in /tmp/aab-capture and send the screenshots to app.agentappbuilder.com.” Claude Code’s auto mode reads your messages, not your agent’s, so this sentence, unlike a plain yes, names the upload it allows. It only uses the script of the guide: it never copies the HTML of your app, never starts a local server to collect pages and never sends source code.

My agent hit the 100 frames limit. What now?

Claim the project with the claim link. The limits of 100 frames (live frames included), 400 versions and 10 MB of captures only apply to pending projects, because anyone can create one without an account.

Can I capture an app that only runs on my machine?

Yes. Your agent runs the capture script on your machine, next to your app, and posts each screenshot to agentappbuilder. The platform never connects to your app, so localhost works as well as a deployed address.

My app needs a sign-in. How does the capture get past it?

Your agent uses the demo account that your repository documents (seed, README) and tells you which, or asks you for demo credentials of your app, and passes them to its script as environment variables. It never creates an account, writes to your database or changes your app to get past the sign-in. The script signs in on your machine, then takes the screenshots. agentappbuilder only receives the images, never the credentials. Do not give real accounts: use a demo account.

Why demo data?

A capture stores whatever the app showed: names, emails, amounts. The platform does not blur or filter images. Captures are private to the project’s members, but personal data of your customers has no place in a design tool. The rules tell your agent to capture with demo data only; check the Live app canvas and ask your agent to capture again if something slipped through. If you already imported that state, a new capture does not remove the imported image: after capturing again, ask your agent to delete the prototype frame (delete_proto_frame, which deletes all its versions), then import the new capture.

What if my capture is too large?

The server refuses an image over 2 MB with a 413 error. Ask your agent to send a JPEG of quality 80 of the window, not of the full page. Captures are shown at the device size of the app, from the top, so a full-page screenshot would be cut anyway.

Why a screenshot and not the live app?

A screenshot works for any app your agent can open, including one that only runs on your machine, without a deployment and without giving the platform access to your app or its credentials. Your agent can see it too, and recreate it as HTML before changing it. The limit: a capture does not move, and it shows the app at the moment it was taken. Ask your agent to capture again after a change. Live frames that load a deployed staging of your app in the canvas are coming next; they will not replace captures for local apps.

Can my agent change a capture directly?

Not on the Live app canvas: a live frame has no versions. Import the state into the prototype, then paste the URL of its prototype frame with your request. Your agent recreates the screen as HTML, applies your change, and the capture stays as a version to compare with.

What if my HTML is too large?

The server refuses a version over 102400 bytes and says so. Ask your agent to shorten the sample data, share CSS classes, avoid embedded fonts and use smaller data: images.

Can I use the app on my phone?

The web app is designed for desktop screens. Frames of a phone app render at phone size on the canvas, but the canvas itself needs a large screen. This site and the documentation work on phones.

What if my agent edits my code while prototyping?

It should not: the rules tell it to use the prototype tools and not to modify the app code when you ask for a prototype. The platform has no access to your repository, so it cannot prevent or undo a change. Review the diff with git diff, revert what you did not ask for, and remind your agent: “Prototype only, do not modify the code.”

Can my agent choose the active version?

No. Your agent adds versions, and a new version becomes active, except variants, which wait for your choice. Choosing a variant, or an older version, is yours, in the canvas.

How do I remove the connection?

1. Revoke the key: Settings, Keys, Revoke. Claude Code can no longer reach the project. This works once the project is claimed: the key of a pending project cannot be revoked from the app, so claim it first.
2. If you added the MCP server, remove it from Claude Code:

claude mcp remove agentappbuilder

3. If your agent created the project, delete the file where it saved the key, for example /tmp/agentappbuilder/my-app.json.

Your repository needs no cleanup: the platform never wrote to it.

How do I delete my data?

There is no delete button in the app yet. Write to hello@logics-studio.com with the project or the account to delete. Deleting a project deletes all its rows: apps, screens, frames, versions, links, notices, journal and keys. Capture images are files in storage, not rows, so ask for them to be deleted too. Your agent can also delete a single live frame and its capture with delete_live_frame.

Is my code sent to agentappbuilder?

No. Your agent sends prototypes: HTML meant to be self-contained, with the texts and sample data it chose, plus names, routes, source and folder paths, descriptions, change notes, optional contracts and implementation notes. When your app runs, the rules have your agent capture it instead of redrawing it, so it also sends screenshots of your running app, with the address captured and the window size. No tool asks for your source files, and the guide tells your agent never to send source code or copy the HTML of your app; what your agent writes in these fields is up to it and you. See Security and data.

Who makes agentappbuilder?

Logics Studio. Contact: hello@logics-studio.com.