IceShore Adoption Checklist for IT Departments

IceShore holds architectures and infrastructure-as-code files, not your business data#

When a team asks to use IceShore, you probably want to check four things first: what data leaves the company, where AI inference happens, how to stop its use, and what to do when you want data deleted. This article answers those four questions, and whether you can take your data out after the contract ends.

IceShore is a configuration management platform that prompts your AI to organize what each system does for the business (its use cases) and hands the result back to that AI. It works by connecting to the AI coding assistants your developers already use, such as Claude Code.

What data IceShore receives, and what it does not#

What IceShore stores is ADL (Architecture Description Language), the text-format data that makes up the content of an architecture, together with the infrastructure-as-code (IaC) files it references (such as Terraform and CloudFormation). ADL contains the following information.

• What the system does for the business (use case names and descriptions).
• Who uses it, and through which entry point.
• Which resources a request passes through, in order.
• What kind of information travels on each path.

IceShore does not receive the following data.

• Application source code. Your AI reads the code locally and writes the ADL, and what it sends to IceShore is the ADL and your infrastructure-as-code files. Those files are stored as they are, so if a file contains function code written inline (such as CloudFormation ZipFile), that code is stored too.
• Your business data. The contents of the data your systems handle, such as customer lists and transaction records, are not sent.
• Credit card information. The payment provider holds it, so it does not reach IceShore.

Data is stored in the AWS Tokyo region (ap-northeast-1). The core data is in Japan. For the information that goes to providers outside Japan and where it goes, see "With service providers" in the Privacy Statement and the separate "Information Sent to Third Parties" page.

In the content IceShore receives from your AI, anything that looks like the value of an access key, JWT, private key, connection string, password, or token is masked automatically before it is stored. A record of each detection is also kept. Masking is not guaranteed to be complete.

Descriptions are free text. Only credentials are masked automatically. Customer names, business partner names, and internal system names are not hidden. Please do not include sensitive information in descriptions or comments.

AI inference happens on the AI you already use#

Both drafting the use cases and answering questions about impact are done by the AI your company subscribes to and uses. IceShore does not perform AI inference. We do not operate any machine learning models. IceShore's role is to prepare the material for that judgment and hand it over. What reaches your AI is the architecture (ADL). Credentials do not.

In other words, adopting IceShore does not add another AI service to the ones you use. If your company has internal rules on AI use, you can judge whether IceShore can be used within them.

The connection uses a standard called MCP (Model Context Protocol). The MCP server logs record the following.

The MCP server logs do not include IP addresses or user identities. The connection correlation ID is an identifier that IceShore issues when an AI client opens a connection, and it does not point to an individual. Separately, the access logs of our content delivery network (Amazon CloudFront) do keep the client IP address, and they are deleted automatically after 90 days.

You can cut a connection at any time#

Connections are authorized with OAuth 2.1 and PKCE. Access tokens last one hour. Refresh tokens last 30 days, and it expires automatically if it goes unused for 30 days. Manual revocation takes effect immediately.

There are three ways to revoke access. All of them are done on screen, and none require contacting IceShore.

• Revoke a single connection. Revoke just that connection from “Connected AI” in the app.
• Revoke all connections at once. You can revoke them all from the same screen. Every AI will need connection approval again the next time it is used.
• Remove a member. When you remove a member from the workspace, that person's connections can no longer read the architectures in that workspace.

You can also show an architecture to people outside the company through a share link that requires no login. The link is a hard-to-guess URL and expires 14 days after it is issued. You can invalidate it at any time before then, and each manual invalidation is recorded in the audit log.

SSO, SAML, and SCIM are not supported. To cut off someone who has left the company, remove them from the workspace.

Deleting architectures and closing accounts#

To delete a single architecture, delete it from that architecture's screen. The confirmation screen states that no one involved will be able to open it, that every share link already issued will stop working, and that it cannot be restored. You must type “Delete” to proceed.

To delete an entire workspace, the confirmation screen shows the number of architectures that will be deleted, and you must type the workspace name to proceed. Neither action can be undone. Note that deleting from the screen makes the architecture inaccessible to everyone. The stored data is not erased from our systems and remains.

To erase a user's data, the user closes their account. Their name and email address are erased, and the account is anonymized. However, architectures that person created remain in the workspace. If the user has lost access and cannot close the account themselves, contact legal@iceshore.ai. After verifying their identity, we will close the account on their behalf.

Contract and billing records remain after deletion or account closure. The audit records themselves accumulate with no time limit on every plan, but how far back you can view them depends on the plan (the most recent 30 days on Free and Team, and 12 months on Business).

Architectures a team created in a personal workspace can be moved to the company workspace. This is a move, not a copy, so they no longer remain in the personal workspace. When an architecture is moved, its share links become invalid.

Your architectures stay with you after the contract ends#

The architecture data (ADL) is text, and Reindeer Technology Pte. Ltd. publishes the language specification and schema under the Apache License 2.0. You can output the use cases you have built up as ADL text through the AI coding assistant you connected, so you can save them locally as files.

You do not need an IceShore contract to read ADL with other tools. After the contract ends, what you have saved locally stays with you.

The specification is public on GitHub. https://github.com/reindeer-project/ArchitectureDescriptionLanguage

Other items for your assessment#

The items this document does not cover are in the following documents.

Please check the recommended environment first. It assumes the latest stable version of Google Chrome and a screen resolution of at least 1920x1080px. Features are limited at widths under 960px.

See it for yourself#

Before you decide, take a look at a sample architecture. No email signup is needed.

https://app.iceshore.ai/workspace/sample/

IceShore is provided by Launchers Co., Ltd. For questions about your assessment, contact support@iceshore.ai.