Skip to content

Privacy Policy

Last updated: November 14, 2026

This Privacy Policy explains how Infragate, LLC (“we”, “us”, or “our”) collects, uses, and discloses information when you use our Services. It also describes your privacy rights and how the law protects you.


Words with capitalized initial letters have meanings defined below. These definitions apply whether they appear in singular or plural form.

For the purposes of this Privacy Policy:

  • Account - a unique profile created for you to access our Services.
  • Affiliate - any entity that controls, is controlled by, or is under common control with a party.
  • Company - refers to Infragate, LLC, located at 1207 Delaware Ave, Suite 679, Wilmington, Delaware 19806, United States.
  • Cookies - small files placed on your device by a website to store browsing data.
  • Device - any device capable of accessing the Services, such as a computer, mobile phone, or tablet.
  • Organization - a workspace that one or more Accounts belong to. Most of our Services are organization-scoped.
  • Personal Data - any information related to an identified or identifiable individual.
  • Services - the Website together with the products described in Section 2: ShareCube, Lanyard, Cairn and Capa.
  • Service Provider - any third party processing data on behalf of the Company.
  • Third-Party Social Media Service - websites or networks that allow login or account creation. We offer Google and GitHub.
  • Usage Data - automatically collected data about your interactions with the Services.
  • Website - refers to infragate.ai, infragate.co, sharecube.io, capa.sh, sharecubeusercontent.io and shrcb.io, and all of their subdomains.
  • You - the individual using the Services, or the organization on whose behalf you use them.

This policy covers all four Infragate products, each of which handles data differently. The product-specific sections are Sections 6 to 9.

ProductWhat it isWhere it runs
ShareCubeEditable documents (“artifacts”), images, video and audio, with files, saved data, comments, sharing and AI chatsharecube.io, app.sharecube.io, artifact pages on *.sharecubeusercontent.io, short links on shrcb.io
LanyardA gateway that connects your AI agents to third-party MCP servers on your behalflanyard.infragate.ai, <project>.lanyard.infragate.ai
CairnTelemetry and analytics for AI agent sessionscairn.infragate.ai, cairn.infragate.co
CapaA command-line tool that runs on your own machine, plus a hosted OAuth relay and package registrycapa.sh, capa.infragate.ai
  • When you sign up yourself, for your own use, we are the controller of your Account data.

  • When you use one of our Services inside an organization that someone else administers (your employer, for example), that organization decides what is collected and why, and it is the controller. We act as its processor. This applies to ShareCube organization content, Lanyard audit records, Cairn telemetry, and any account created through your organization’s single sign-on.

    If your organization is the controller, direct access and deletion requests to it first. We will help it respond. That processing is governed by our Data Processing Agreement, which applies automatically to every organization and does not need to be requested or signed.

  • When someone collects data from other people through a ShareCube artifact, for example an RSVP page or a form, the artifact’s owner, or the organization it belongs to, is the controller of what is collected, and we store it on their behalf as their processor. That’s true for a personal organization too. See “Data visitors save in an artifact” in Section 6.


CategoryWhat
AccountEmail address, first and last name, profile photo
ProfileCompany or organization name, job title or role
OrganizationOrganization name, logo, domain verification records, membership and roles
BillingBilling address and tax ID, collected at checkout. Card details go directly to Stripe and never reach our servers
InvitationsThe email address of anyone you invite who is not yet a user. Stored for 7 days, then deleted
SupportWhatever you put in an email to us
Content reportsIf you report an artifact, the category, your description and, if you give them, your name, email address and organization. A copyright report requires them. You don’t need an Account to report

If your organization uses single sign-on, its identity provider sends us your email address, first and last name, and group memberships, and your Account is created automatically. Administrators of an SSO organization manage your profile; you cannot edit it yourself.

Organization administrators can also upload a reporting hierarchy (a CSV naming each person’s manager), which feeds organization charts and usage analytics. The uploaded file is deleted after 7 days; the hierarchy it describes is kept.

Administrators control session length, domain verification, membership, and, in Cairn, per-person usage reporting. Where your employer controls the Account, your employer is the controller for this data.

When you join a waitlist through a form on our Website we collect your email address and name and, optionally, your company, role and the product you are interested in. We use it for one thing: to tell you when access to that product opens. Joining a waitlist does not subscribe you to a mailing list, and we do not send marketing from it. We keep a waitlist entry for one year, and you can ask us to remove your details sooner at privacy@infragate.co.

Waitlist sign-ups are also posted to our internal Slack workspace so the team sees them.

Collected automatically, and may include IP address, browser type and version, pages visited, visit time and duration, device identifiers and diagnostic data. On a mobile device this may include the model and operating system.

If you sign in through Google or GitHub we receive the profile data those providers release for sign-in: typically your name, email address and avatar. For sign-in we request only openid email profile (Google) and read:user user:email (GitHub); we do not request access to your repositories.

This is separate from connecting a GitHub or GitLab account to Capa, which does request repository access. See Section 9.

We keep an internal reporting copy of user id, email and name, and of organization membership, so we can measure product usage.

We also record product events on our servers, such as an artifact being created, viewed or shared, a comment being added, or an email being sent. Each event carries ids and coarse values only (an artifact’s type and size range, say), never content, titles, names or email addresses. A view of a public artifact by someone who isn’t signed in is recorded without any id for the viewer, and a view through an embed on another website is recorded the same way, marked as an embed. We also record how you first arrived, for example from ChatGPT, Claude or our own site, and which AI client made a call over MCP.

None of this uses cookies or runs in your browser, and none of it is shared outside the Company. We keep events for 25 months, and delete yours when you delete your Account. To opt out, email privacy@infragate.co: we delete your events and your recorded sign-up origin, and record nothing new about you.


Our public websites use Google Analytics, and only if you agree. That covers infragate.ai, sharecube.io, capa.sh, blog.infragate.ai and docs.infragate.ai. The consoles, the CLI and everything you sign in to run no cookie or browser analytics (the server-side product events in Section 3 are separate), and none of our sites run advertising pixels, session recorders or a tag manager.

On those public sites, a banner asks before any analytics cookie is set:

  • Until you choose, or if you decline, Google Analytics runs in consent mode with cookies off. Google receives a cookieless ping for each page view (the page, referrer, browser type and your IP address, which Google does not store) and we see only rough, modelled totals.
  • If you accept, Google Analytics sets _ga cookies for up to 2 years so it can tell a returning visitor from a new one. We use this to see which pages get read. Advertising and ad personalization signals stay off either way.
  • Your answer is kept in an analytics_consent cookie for 180 days. Change it any time by opening infragate.ai, sharecube.io or capa.sh with #cookie-settings at the end of the address. Declining deletes the _ga cookies.

Google processes this data under Google’s privacy policy, in the United States.

Other cookies and storage:

  • Session cookies set by our identity service, to keep you signed in. These are HttpOnly and are required, because authentication will not work without them.
  • Local storage and session storage in the ShareCube, Lanyard and Cairn consoles, to hold your sign-in tokens and preferences such as theme, recently viewed items and navigation state. This stays in your browser.
  • Stripe sets its own fraud-prevention cookies on pages where you enter or manage a payment method. These are governed by Stripe’s privacy policy.
  • Artifact pages are served from sharecubeusercontent.io, on a separate sandboxed subdomain for each artifact. We set no cookies there, and a page can’t read your ShareCube sign-in.
  • Short links on shrcb.io only redirect to ShareCube. We set no cookies there.
  • Embedded players. When public media is embedded in another website, the player in it comes from ShareCube. We set no cookies in it, and it sends no ShareCube sign-in.

PurposeLegal basis (UK/EU GDPR)
Providing, maintaining and supporting the Services; managing your Account and access rightsPerformance of a contract
Taking payment, preventing fraud, and meeting tax and accounting obligationsContract; legal obligation
Securing the Services, investigating abuse, enforcing rate limits and plan quotasLegitimate interests
Screening images and video for child sexual abuse material and other content our Terms prohibit, and holding content while we review it (Section 6)Legitimate interests, ours and the public’s, in keeping abuse material off the Services
Reporting apparent child sexual abuse material to the National Center for Missing & Exploited Children, and preserving it, as United States law requires of us (Section 6)Legitimate interests. The duty to report comes from United States law
Diagnosing faults and improving the ServicesLegitimate interests
Product usage analytics and sign-up origin (Section 3)Legitimate interests
Storing data visitors save in an artifact, on its owner’s behalf (Section 6)Performance of a contract with the artifact’s owner
Service and security announcements you cannot opt out ofContract; legitimate interests
Product emails about our own products, such as getting-started tips and product news, which you choose at sign-up and can unsubscribe from in every emailYour consent, where the law of the country you sign up from requires it (the switch starts off). Elsewhere, legitimate interests, sent under the local email law, such as CAN-SPAM in the US
Marketing email about anything other than our own products, if we ever send itConsent, withdrawable at any time
Responding to legal requests, defending claims, and business transfersLegal obligation; legitimate interests

Product emails. We send customers emails about our own products: today a short series of getting-started tips after you sign up, and in future product news. We treat these as marketing email. When you sign up, a switch lets you choose whether to receive them. Where it starts depends on the country you sign up from. Where its law requires your consent first, for example in the European Economic Area, the UK, Switzerland, Canada, Australia, Israel and Turkey, the switch starts off and we send them only if you turn it on. We also start it off in some countries where that is the safer reading, such as India and Japan, and when we cannot tell where you are. Elsewhere it starts on and you can turn it off. Each product email has a link to unsubscribe. We record your choice, and if you turn it on, when, from which country and the words the switch showed, so we can show your consent. If your Account existed before this policy took effect, we do not send you product emails unless you turn the switch on when we next ask you to accept these terms. Service and security announcements, invitations and notifications you set up are not product emails and are not affected.

We do not sell personal information, and we do not share it for cross-context behavioural advertising.

We do not use your content to train machine learning models. That covers ShareCube artifacts, their files and saved data, and comments, Cairn telemetry, and Lanyard tool call data. Where content is sent to an AI model to answer a request you made, it is sent for that request only. See Sections 6 and 8.


ShareCube stores the documents you and your agents create, called artifacts, together with their files, saved data, version history, comments, folders, project names and organization membership. Artifacts may contain whatever you choose to put in them.

To operate the product we also process artifact content: we extract text to build a full-text search index and a per-organization concept graph, and we generate preview cards and summaries for publicly shared items. All of this runs on our own infrastructure.

There are three visibility levels:

  • Private - visible only to members of the project it belongs to.
  • Organization - visible to everyone in your organization.
  • Public - readable by anyone holding the link, without signing in, and it may be indexed by search engines. A signed-in user from another organization may also be able to comment on a public artifact.

Please do not put sensitive personal data in an artifact you intend to share publicly.

A shared artifact is user-authored HTML or Markdown and runs in a sandboxed frame. It can load resources from, or send requests to, third parties we neither control nor vet. Treat an artifact from someone you do not trust the way you would treat any other untrusted web page.

You can add files to an artifact, such as images, video, PDFs, data files and fonts, by uploading them or by giving us a link that we fetch for you. When we fetch a link, the site it points to sees a request from our servers, not from you.

We remove location, camera and other metadata from JPEG, PNG, WebP and GIF images before we store them, because a photo is shown to everyone the artifact is shared with. Other files, such as AVIF images, video, audio and PDFs, are stored as you upload them, so remove anything you don’t want shared before you add them.

Files are part of the artifact. They’re deleted when the artifact is deleted. A file you replace or remove stays stored for 90 days after it leaves the current version, so that an earlier version can be restored, and is then deleted. During those 90 days it doesn’t count toward your plan’s storage.

An artifact can itself be an image, a set of images, a video or an audio file, and any artifact can hold images and video as files. We make processed copies of them on our own infrastructure: smaller versions of an image; for a video, versions that play in a browser, a poster image, a sheet of small preview frames and a waveform of its sound; and for an audio artifact, a version that plays in a browser, a waveform and a picture of that waveform for previews. We also read basic facts from the file, such as its dimensions, its length and how loud it is.

  • Processed copies carry no metadata. They also carry no cover art from an audio file. Originals follow the rule under “Files” above: location and camera data is removed from JPEG, PNG, WebP and GIF originals, while video, audio and AVIF originals are stored as you upload them and can still say where and when they were recorded, or carry other details and cover art.
  • Who gets the original. Project members, and people invited to the project or the artifact, can open the original. Someone who reaches an image, video or audio artifact through its organization link or public link gets processed copies only. The artifact’s Owner can turn on downloads for link visitors, which hands them the original of an image or video as stored. That setting is off by default. An image, video or audio file placed inside an ordinary page is different: it is served as stored to everyone who can open the page.
  • Stems and captions. A video or audio artifact can also hold up to 4 stems, which are extra audio tracks such as a voice or a music track, and caption files you upload. Stems are for the project: people who reach the artifact through an organization link or public link can’t see or play them. Captions are shown to everyone who can open the artifact. Comments can be pinned to a moment, a caption line or a stem.
  • People in your media. Images, video and audio often show people or carry their voices. We store and deliver them as your content. We do not identify anyone in them: we run no facial or voice recognition, create no face or voice templates, don’t transcribe speech, and don’t work out who appears in a file.
  • Embeds. An image, video or audio artifact with a public link can be embedded in other websites, and anyone with the link can embed it. When someone opens a page with an embed, their browser loads the player and the media from us, so we receive their IP address, their browser and, usually, the address of the site the embed is on. That goes into our edge logs like any other request. An embed stops showing the artifact once it is no longer public or is deleted, though a page already open can keep loading its files for a few hours. A hold stops it at once.
  • After you remove someone. A link to an artifact’s files that a person has already opened keeps working for up to 6 hours, so someone removed from a project can keep loading that version’s files for up to 6 hours afterwards. A hold stops delivery at once.

We screen images and video for child sexual abuse material and for other content our Terms of Use prohibit. Screening is automated and runs without anyone asking for it.

  • Images are screened when they are added. A reduced copy goes to Amazon Rekognition content moderation, an AWS service, which answers with labels for what the picture appears to show, such as explicit nudity or graphic violence. For an animated image we send up to 8 of its frames.
  • Videos are screened the same way, from a reduced copy, the first time the artifact becomes reachable by someone outside its project: through an organization or public link, or through an invitation to someone from another organization. Until that finishes, those people see “Checking this video” instead of the video. Project members are not held back.
  • Audio is not screened. We don’t analyse sound, in audio or in video, and people who reach audio through a link get it at once. An audio artifact whose file carries cover art can’t be put on an organization or public link until the art is removed, because that picture would not be screened.
  • Known abuse material. For images and video we also compute a perceptual fingerprint (PDQ for images, TMK+PDQF for video). We compare these fingerprints with lists of known child sexual abuse material from child safety organizations, as we obtain access to those lists. This runs on our own infrastructure, and the file is not sent to anyone for it.
  • This is not biometric identification. Moderation labels describe what an image shows, not who is in it, and a perceptual fingerprint describes a whole picture or video, not a person.

What happens on a result. If moderation is confident that an image or video shows explicit, violent or disturbing content or hate symbols, or if an image that is already shared outside its project can’t be checked, the artifact is put on hold automatically, before a person has looked at it. It stops being served, its owner sees that it is held, and our staff then review it and release it if the result was wrong. A video that can’t be checked is usually not put on hold, but people outside its project keep seeing “Checking this video” instead of it. If you think a hold is a mistake, email privacy@infragate.co and a person will look at it.

If a fingerprint matches known child sexual abuse material, the artifact is held and we keep a copy of the original apart from the artifact. Whenever we find apparent child sexual abuse material, through screening, a report or otherwise, we report it to the National Center for Missing & Exploited Children (NCMEC) through its CyberTipline and preserve it, as 18 U.S.C. § 2258A requires. A report can include the file and what we hold about the upload and the account behind it, such as the email address and IP address. We may not be able to tell the account owner about a report.

Screening is not exhaustive. It covers images and video stored as files. It does not cover, for example, audio, pictures embedded inside a page’s own content or inside an audio file, or other kinds of file. We are not required to monitor content, and screening is no promise that everything on the Services has been checked.

An artifact’s owner can let the page save data, such as RSVPs, votes, sign-ups or form answers. If the artifact is public, that can include visitors who aren’t signed in to ShareCube.

  • Who’s responsible for it. The artifact’s owner, or the organization it belongs to, decides what the page collects and why. They’re the controller of that data, and we store it on their behalf as their processor. Only the people the owner’s rules allow can see it.
  • What we store. What the page sends, and when. For a visitor who isn’t signed in we attach no account, name or IP address. We use a hashed form of the IP address that changes every day, only to limit abuse, and discard it within about a day. For a signed-in person, the page only ever sees an id that is different on every artifact. We keep the link from their own entries to their Account, so we can delete those entries when they leave the organization or delete their Account.
  • Your rights. If you gave data to an artifact, ask its owner to show it to you, correct it or delete it. You can also email privacy@infragate.co and we’ll pass your request on, but for a visitor who wasn’t signed in we usually can’t tell which entries are yours.
  • When it’s deleted. All of an artifact’s saved data is deleted with the artifact. Data a signed-in person saved under their own name is deleted when they leave the organization or delete their Account. Data they added to shared parts of a page, like a shared list, stays, the same as their comments do.

Saved data is not used for analytics. It goes to an AI model only when someone with access to it asks Cubie, or an AI client they connect over MCP, to work with it, and then only for that request, as described under “Connecting an AI client” below and in Section 8.

ShareCube’s chat sends the artifact content or selection you are asking about, together with your conversation, to a large language model through AWS Bedrock, so the model can answer. The models currently configured are Anthropic Claude (Sonnet and Haiku) and OpenAI’s gpt-oss, all accessed inside AWS. Chat may also use an AWS-operated web search connector.

Cubie can’t watch a video, so when you ask about one it can fetch still frames from it, and those go to the model the same way, for that request only. Cubie can’t listen to audio either. For audio and video it can read the caption files and the waveform, which go to the model the same way, but the sound itself is not sent to a model.

Chat transcripts are retained for 90 days. Chat content is not used to train models.

Version history, notifications and presence

Section titled “Version history, notifications and presence”
  • Previous versions of an artifact are kept for 90 days and then deleted. Files follow their own rule, described under “Files” above.
  • Notifications and digest emails include the artifact name and up to 280 characters of the comment that triggered them.
  • Artifacts and comments you author appear on a profile timeline visible to your organization. You can make your timeline private in settings.
  • Each comment keeps the name its author had when they wrote it, so a discussion still reads correctly after someone leaves the organization. While the author is a member we show their current name instead. The stored name is part of the organization’s content and is deleted with the comment, the artifact or the organization.
  • Edge and API access logs record the IP address, browser, page or endpoint, and outcome of each request, and which signed-in user made it. They are kept for 30 days and used for security, for answering access requests, and to measure video and audio delivery per account. For that measurement we read how many bytes of video and audio each organization’s artifacts delivered and keep only a monthly total per organization, with no IP addresses and no record of who watched.
  • While you have an artifact open, other viewers can see that you are viewing it.

When you connect an AI client such as Cursor, Claude or ChatGPT to ShareCube over MCP, the content it fetches goes to that provider under that provider’s terms, not ours. That includes pictures: a client can ask for a preview of an image and for still frames of a video, and receives them as images. For audio and video it can also read caption files and the waveform, and get a link that plays the audio. Stems are given only to clients connected by someone in the project or invited to it. We keep a still frame we have drawn for 7 days so it doesn’t have to be drawn again. You can revoke a connection in Settings. MCP access tokens last 1 hour and refresh tokens 30 days; API keys you create do not expire until you delete them.


Lanyard stores the credentials for the third-party services you connect to it: OAuth access and refresh tokens, and API tokens you paste in.

These are envelope-encrypted with AES-256-GCM under a per-record key from a multi-region AWS KMS key with rotation enabled and a restrictive key policy. Tokens you paste are additionally encrypted in your browser before they reach us. Stored alongside the encrypted credential, in plaintext, are the owner’s email address, the server URL, the issuer, the granted scopes, and the text of the last refresh error.

Credentials are kept until you delete the connection or leave the organization. Deleted records remain in point-in-time database backups for up to 35 days.

Deleting a connection in Lanyard does not revoke the token with the provider. To fully revoke access, also remove the authorization in the provider’s own settings.

The content of your tool calls, meaning the arguments and the results, is not stored or logged. Our logs write only an allowlist of fields, and every value passes through a secret scrubber.

We do record metadata about each call, because the product’s audit log, security controls, rate limiting and billing depend on it: time, organization, project, user, API key id, the server host and tool name, outcome, error and upstream status, latency, and the client IP address and user agent. Aggregates over this metadata power the usage dashboards.

Audit history is retained by plan: 7 days on Free, 30 days on Pro and Team, 365 days on Enterprise. Recent activity is kept for 30 days. Edge access logs, which include IP addresses, are kept for 30 days.

When one of your agents calls a tool, the arguments and the credential go to whichever upstream MCP server you chose to connect. Those servers are third parties acting at your direction, not our service providers, and what they do with the data is governed by their terms, not ours. We register with providers as the OAuth client “Lanyard by Infragate”.

We email the connection owner about expiring credentials and changed tools.

The Lanyard console loads Google Fonts and uses Google’s favicon service to show an icon for each connected server. The favicon request tells Google your IP address and the domain of the server you connected. There is no analytics in the console.


Cairn is agent observability. It exists to retain and analyse the telemetry you send it, so it is the Service that holds the most sensitive data.

Cairn ingests OpenTelemetry spans and events from your agents and stores their attributes, including attributes we do not recognise. Depending on how you configure your agent, those attributes can include user prompts, assistant responses, and tool inputs and results. Our setup instructions show how to switch that on (OTEL_LOG_USER_PROMPTS, OTEL_LOG_ASSISTANT_RESPONSES, OTEL_LOG_TOOL_DETAILS).

You choose what to send. If you do not want prompt and response content in Cairn, do not enable those flags.

Every attribute is screened for personal data with Amazon Comprehend before it is stored. This is a best-effort filter, not a guarantee: it analyses English only, it is not a secrets scanner, and an organization administrator can turn it off. Anything stored while redaction is off stays unredacted. Do not send secrets, payment card data, health data or special-category personal data to Cairn.

Cairn telemetry currently has no automatic expiry: sessions, spans, events and metrics are retained until you ask us to delete them. Queue buffers holding not-yet-redacted payloads clear within about 4 days. We are working on configurable retention; until then, email privacy@infragate.co to have data deleted.

  • Session evaluation is opt-in. When you enable it, rendered sessions, which may include prompts, replies and commands, are sent to Amazon Nova Lite through AWS Bedrock for scoring.
  • Per-person profiles. Organization administrators can see per-person reporting: name, email, cost, usage patterns, and where sessions stall. If you deploy Cairn across a team, this is monitoring of your staff, and giving them notice and having a lawful basis for it is your responsibility, not ours.

Cairn telemetry is not used to train models.

The Cairn console loads Google Fonts. There is no analytics in the console.


Capa is a command-line tool that runs on your own machine.

  • No telemetry. The CLI sends us no analytics, usage data or crash reports.
  • Local activity traces. Tool arguments and result previews for the last 10,000 calls per project are stored in ~/.capa/capa.db on your machine. They are never uploaded.
  • Update check. Every command checks api.github.com for the latest release. That request is unauthenticated, but it tells GitHub your IP address, user agent and the time.
  • Local web UI. Capa’s local web interface loads Google Fonts.

capa auth runs the OAuth exchange through an edge function at capa.infragate.ai/auth, which exists so the OAuth client secrets stay out of the CLI. It stores nothing: it swaps the authorization code for tokens and redirects them straight to the CLI listening on your own machine. Token refresh is proxied the same way. If you supply a personal access token with --access-token instead, it never touches our infrastructure at all.

The access Capa requests is broader than sign-in:

ProviderScopesWhat that means
GitHubrepoRead and write access to all of your public and private repositories
GitLabread_api, read_repositoryRead access to the API and to repositories

You can revoke this at any time in your GitHub or GitLab account settings.

Capa installs and runs skills, hooks, MCP servers and plugins from GitHub, GitLab, registries and URLs you point it at. That code runs on your machine with your permissions, and we do not review it. Install only what you trust.

For content published to the Capa Registry we store the publisher identity in the package manifest (name and email), stars, and issue authorship. Package listings and issue authorship are public. Registry content is served publicly and may be mirrored into catalogs.


These process data on our behalf, under contract:

ProviderWhat they processWhere
Amazon Web ServicesHosting, storage, databases, email (SES), AI models (Bedrock), image and video screening (Rekognition), personal-data screening (Comprehend)United States, with edge processing at AWS locations worldwide
StripeSubscription and credit payments, billing address, tax ID, card detailsUnited States and Ireland
SlackWaitlist sign-up and content report notifications to our internal workspaceUnited States
GoogleWeb fonts, and favicons in the Lanyard console. Receives visitor IP addressesGlobal CDN
Google, GitHubSign-in, when you choose social loginUnited States

We will update this list when it changes. The same list is Schedule 3 of the Data Processing Agreement, where it serves as the contractual subprocessor schedule, and changes there carry 30 days’ notice and a right to object.

Separately from the above, data goes to third parties because you told it to. These act at your direction under their own terms, and we are not responsible for them:

  • Upstream MCP servers you connect through Lanyard.
  • AI clients such as Cursor, Claude or ChatGPT that you connect to ShareCube.
  • Resources loaded by a shared ShareCube artifact.
  • GitHub and GitLab, when you connect them to Capa.

We may also disclose data:

  • In a merger, acquisition or sale. You will be notified before a different policy applies to your data.
  • To law enforcement or public authorities where required by law.
  • To the National Center for Missing & Exploited Children (NCMEC), a United States non-profit, when we find apparent child sexual abuse material. United States law (18 U.S.C. § 2258A) requires us to report it, and NCMEC may pass a report on to law enforcement in the United States or elsewhere. See “Automated screening of images and video” in Section 6.
  • Where necessary to comply with laws, protect our rights and property, prevent fraud or wrongdoing, protect user or public safety, or defend against legal claims.
  • With your consent, for anything else.

DataRetention
Account and organization recordsFor as long as the Account exists. Deleting the Account removes them immediately
Sessions30 days
Authorization codes5 minutes
Pending invitations, and the invitee’s email7 days
Removed organization memberships30 days
Uploaded reporting hierarchy files7 days
ShareCube artifacts, comments, folders and projectsUntil you delete them; deletion purges the content
ShareCube filesUntil the artifact is deleted. A file replaced or removed from it is deleted 90 days after it leaves the current version, see Section 6
ShareCube processed copies of images, video and audioDeleted with the file they were made from
ShareCube still frames drawn from a video7 days
ShareCube monthly video and audio delivery totals per organizationAbout 13 months
Material preserved because we found apparent child sexual abuse material, and the record of it1 year, or longer where the law requires or allows it. Deleting the artifact or the Account does not remove it
ShareCube saved dataUntil the artifact is deleted. A signed-in person’s own entries go when they leave the organization or delete their Account, see Section 6
ShareCube previous versions90 days
ShareCube chat transcripts90 days
ShareCube edge and API access logs30 days
Lanyard credentialsUntil the connection is deleted or you leave the organization
Lanyard audit history7 days (Free), 30 days (Pro, Team), 365 days (Enterprise)
Lanyard recent activity and edge access logs30 days
Cairn telemetryNo automatic expiry. Deleted on request, see Section 8
Product analytics events25 months, or until you delete your Account or opt out, see Section 3
Waitlist entries1 year from sign-up, or sooner if you ask us to remove them
Billing records and invoicesKept by us and by Stripe as long as tax and accounting law requires
Operational logsVaries by component, from 14 days to indefinite. We are standardising this
Database backupsDeleted records remain restorable for up to 35 days
Content reportsKept while we review them and as long as we need a record of the decision

Where we say “deleted on request”, see Section 13.

Content on hold. If we put an artifact on hold while we review a report or a screening result (Terms of Use 4.4), it stops being served and can’t be deleted until the hold is resolved, and we keep it, its files and its saved data until then, even if you try to delete it. After a delete, a link someone already opened can keep working from cache for up to a few hours. After you remove someone from a project, a link to an artifact’s files that they already opened keeps working for up to 6 hours.


Our infrastructure runs in the United States (AWS us-east-1). Some processing happens at AWS edge locations worldwide: ShareCube uploads are handled at the CloudFront location nearest you, and Lanyard and Capa run edge functions. Logs from that processing stay in the region where it ran.

If you are in the UK, EU or another jurisdiction that restricts international transfers, your data will be transferred to the United States. That transfer is covered by our Data Processing Agreement, which incorporates the EU Standard Contractual Clauses (Module Two) and the UK Addendum, and sets out the equivalent position for Switzerland. It applies automatically; you do not need to ask us for it. If your procurement process needs a countersigned copy, email privacy@infragate.co.


Wherever you are, you can ask us to:

  • Access the personal data we hold about you, and receive a copy in a portable format
  • Correct anything inaccurate
  • Delete your data
  • Object to or restrict processing based on our legitimate interests
  • Object to direct marketing, including our product emails, at any time. We will stop, with no questions asked
  • Withdraw consent to marketing at any time, without affecting anything done before

Email privacy@infragate.co. We will respond within 30 days, and will tell you if we need longer. We will not charge you or treat you differently for exercising a right.

If your organization administers your Account, ask your organization first. We act on its instructions for that data and will refer your request to it. The same goes for data you gave to someone else’s ShareCube artifact: ask the artifact’s owner first, see Section 6. This does not apply to product emails: we decide how those are used, so contact us directly.

How to delete your Account. Go to the Security page of your account settings and choose Delete my account. Your personal organization must be the only one you belong to, so leave or delete any other organization first; if you own one, delete it or transfer ownership and leave. Deletion is immediate, removes your profile and your personal organization along with the Account, and cannot be undone.

If your Account is administered by your organization through single sign-on, you will not see the option, because the Account belongs to that organization. Ask an administrator there.

Two things survive: invoices and payment records we and Stripe must keep for tax and accounting, and database backups, from which deleted records age out within 35 days. If you would rather we did it for you, email privacy@infragate.co from the address on the Account and we will do it within 30 days.

In the UK or EU, you also have the right to complain to your data protection supervisory authority.

In California and other US states with privacy laws, the rights above cover the access, correction, deletion and portability rights those laws give you. We do not sell personal information and do not share it for cross-context behavioural advertising, so there is nothing to opt out of; we honour Global Privacy Control signals as a matter of course. We do not use personal data to make decisions that produce legal or similarly significant effects about you. Automated screening can put an artifact on hold before a person has looked at it (Section 6). Our staff review those holds, and you can ask us to look again.


Infragate is built with security as a priority.

  • All traffic is encrypted in transit, and data is stored with server-side encryption at rest.
  • Lanyard credentials are additionally envelope-encrypted with per-record keys from a rotating KMS key (Section 7).
  • Access to production data is governed by IAM policies on the principle of least privilege.
  • The content of tool calls through the Lanyard gateway is not stored or logged. Call metadata is; see Section 7 for exactly what. ShareCube artifacts and Cairn telemetry are stored by design, because showing them back to you is the product.

No method of transmission or storage is completely secure, and we do not claim otherwise.

Reporting a vulnerability. Email security@infragate.co. Please give us a reasonable chance to fix an issue before disclosing it publicly.

Breach notification. If a breach affects your personal data and is likely to result in a risk to your rights, we will notify you and the relevant supervisory authority without undue delay, and within 72 hours of becoming aware of it where the law requires that.


Our Services are not directed at children. You must be at least 18 to use them, and we do not knowingly collect personal data from anyone under 13. If you believe a child has given us personal data, contact us and we will remove it.


Our Services may link to sites we do not operate. We are not responsible for their privacy practices. Please review their policies individually.


We may update this Privacy Policy. The “Last updated” date at the top changes whenever we do, so it always names the version you are reading.

Where a change is material, we ask you to accept the updated policy the next time you sign in, and we record which version you accepted and when. Signing in requires accepting it.

We do not rely on a notice email to tell you about a change, so the date above and that prompt are how you will know. You can read the current policy at any time without signing in.