Kitestring Privacy Policy

Last updated: September 5, 2026

This policy covers the Kitestring desktop app for macOS and the Kitestring dashboard.

Overview

Kitestring is a workplace AI analytics and coaching tool deployed by organizations to their teams. It gives your organization’s leadership an aggregate picture of how AI is being adopted, and gives you, individually, private coaching that helps you get more out of the AI tools you already use.

Kitestring is provided to you by your organization. Your organization is the controller of the data Kitestring produces; Kitestring operates as its processor.

This policy explains what Kitestring captures, what happens to it, who can see it, and how long it is kept. The short version:

• Kitestring captures the text of the prompts you write to supported AI tools, along with usage metadata. It does not capture the AI’s responses, only their length.
• Your prompt text exists to power your own coaching. In the product, it is visible only to you. Your organization sees aggregated analytics: categories, trends, usage and cost totals. It does not see what you typed.
• Prompt text is encrypted in transit and at rest and is never used to train models. It is kept while your account is active, or for a shorter retention window where your organization sets one.

What Kitestring captures, and from where

The Kitestring desktop agent (macOS) reads your conversations with supported AI apps using the operating system’s accessibility interface. It is restricted to a fixed allowlist of AI applications and AI websites: the Claude desktop app, the ChatGPT desktop app, and claude.ai (including Claude’s design tool) and chatgpt.com in Chrome and Safari, gemini.google.com, www.perplexity.ai and chat.deepseek.com in Chrome, and the Gemini side panel inside Google Docs and Google Sheets in Chrome (only the panel’s own conversation is read; the document or spreadsheet itself, its outline, cells and any screen-reader text are never read). It cannot and does not read any other application, window, website, or browser tab. The one narrow exception is material you paste into a supported AI tool, described under “Permissions” below.

Earlier versions of Kitestring included a browser extension. It has been retired and is no longer distributed; the desktop agent is the only capture surface.

From the desktop agent, Kitestring receives:

• The text of your prompts, as you wrote them. This is the material your personal coaching is built from: it is what lets Kitestring show you how to improve a specific ask rather than offer generic tips. This includes material you paste into a supported AI tool as part of an ask; where pasted text cannot be read, Kitestring records only that a paste of a given length happened.
• Usage metadata: which AI tool and model was used, conversation identifiers, timestamps, measured active time in supported tools, whether a conversation used features like projects or produced an artifact, and the titles of artifacts (documents, code, designs) the AI produced.
• A character count of each AI response. The content of responses is never transmitted.
• Classification labels derived from your prompts: the type of work (for example writing, research, analysis) and its subject area.
• Your work identity: you sign in with your work account, and Kitestring receives your work email address to associate usage with your account within your organization. Your password is handled by your identity provider (for example Google) and is never sent to or stored by Kitestring.

Who can see your prompt text

You. Your prompts power your personal coaching view, which can quote your own asks back to you with concrete suggestions. That view is yours alone.

Not your organization. The product does not show your prompt content to your organization’s administrators, managers, or leadership. What your organization sees is aggregated: work-type and subject categories, adoption and usage trends, active-time and cost totals. Requests your organization makes about your data as its controller (see “Your choices and rights”) are handled through Kitestring as processor, not through any content view in the product.

Kitestring personnel, only exceptionally. Kitestring staff do not access prompt content in the ordinary course of operating the service. Access happens only where strictly necessary: support you have asked for, investigation of a security incident, or a legal obligation. Any such access is recorded, and that record is available to your organization’s administrators.

Automated processing. Your prompt text is processed by Kitestring’s systems to classify it and to generate your coaching. Parts of this processing use AI model providers (for example Anthropic) as subprocessors, under terms that prohibit them from training models on your data.

How your prompt text is protected

• Encrypted in transit (TLS) between your device and Kitestring, and between Kitestring and its subprocessors.
• Encrypted at rest (AES-256) in Kitestring’s production data store.
• Access-controlled: tenant isolation separates each customer’s data, and the individual-only visibility described above is enforced server-side, not just hidden in the interface.
• Never used for training. Kitestring does not train models on your data, and its AI subprocessors are contractually prohibited from doing so.
• Retained under a defined policy: see “Data retention” below.

Kitestring is currently undergoing SOC 1 and SOC 2 examinations.

What Kitestring does not do

• It does not record your AI responses beyond their length.
• It does not show your prompt content to your employer; your organization sees aggregated analytics only.
• It does not read any application or website outside the supported AI tools listed above.
• It does not monitor your clipboard. The desktop agent reads the clipboard only when you paste into a supported AI tool, uses the text only when it matches that paste, and immediately discards anything else.
• It does not train models on your data, and does not permit its subprocessors to.
• It does not sell your data.
• It does not use anything captured from your AI tools, or anything from inside the product, for advertising. The only advertising measurement Kitestring does is on its public website, described under Website and product analytics.

How the information is used

The signals Kitestring derives are used for two things: to give your organization’s leadership visibility into AI adoption, usage patterns, and spend at an aggregate level, and to power a private coaching experience that helps you improve how you work with AI, grounded in your own real usage. People administering Kitestring for your organization see patterns, classifications, and totals, not the content of what you typed.

Website and product analytics

Kitestring uses Google Analytics on its website and in the dashboard to understand how people find the product, install it, and get to their first coaching card. This is product measurement, not tracking of what you do inside AI tools. It is entirely separate from the capture described above, and none of the material Kitestring captures from AI tools is ever sent to Google.

What is measured: pages viewed, and a small set of named actions such as starting a download, creating an account, receiving a first coaching card, and starting or completing a subscription. Google sets an identifier in your browser to recognize a returning visit. Dashboard pages that identify a person in their address are reported in a generic form (for example /people/[id]) so that no per-person identifier reaches Google.

Some of these measurements are sent by Kitestring’s servers rather than by your browser, including installing the desktop app, granting the Accessibility permission, capture starting for the first time, and app updates. The desktop app never contacts Google. So that one person’s journey reads as one journey rather than several disconnected ones, Kitestring passes the browser’s Google identifier to the desktop app when you sign in, and includes a random account identifier with these measurements. That identifier is generated by Kitestring and is not your name, your email address, or anything derived from them.

Kitestring also uses the Meta Pixel on its public website (www.getkitestring.com) to measure whether its advertising on Meta platforms leads to downloads and subscriptions. The pixel runs only on the website, never in the dashboard or the desktop app. It reports pages viewed on the website and a small set of website actions: starting a download, requesting a download link by email, and opening the demo booking form. Meta sets an identifier in your browser and, if you arrived from a Meta ad, records that you did.

To measure the same journey end to end, Kitestring’s servers report four later milestones to Meta: creating an account, receiving a first coaching card, starting a checkout, and completing a subscription (with its price). These reports carry the same Meta browser identifier, plus the internet address and browser type of the session that started the checkout, and nothing else: no email address, no name, no account identifier, and nothing about how you use AI. Meta may use this to measure and improve the delivery of Kitestring’s ads, as described in Meta’s data policy. Nothing captured from your AI tools is ever sent to Meta.

Kitestring also uses OpenAI’s ChatGPT Ads measurement pixel on its public website, for the same purpose: to measure whether its advertising in ChatGPT leads to downloads. It runs only on the website, never in the dashboard or the desktop app, and reports the same things as the Meta Pixel: pages viewed on the website, starting a download, requesting a download link by email, and opening the demo booking form. OpenAI sets an identifier in your browser and, if you arrived from a ChatGPT ad, records that you did. Nothing is reported from Kitestring’s servers to OpenAI’s ads service, and nothing captured from your AI tools is ever sent to it.

Never included in analytics: your prompt text, AI responses, conversation or project names, file names, your email address, or your organization’s name. Google Analytics measurements are sent with personalized advertising disabled and are not used for advertising. The Meta Pixel and the ChatGPT Ads pixel are used only to measure Kitestring’s own advertising. Analytics data is not sold. Browser settings that block cookies or analytics scripts prevent this collection, and Kitestring works normally without it.

Permissions

Desktop agent (macOS): the agent asks for the system Accessibility permission. This is what allows it to read the on-screen text of the allowlisted AI apps and sites listed above, and it is used for nothing else. The agent also uses network access to send the captured signals to Kitestring, and local storage to queue signals while you are offline so none are lost.

Some AI tools display a long paste as a collapsed attachment whose full text is not readable on screen. To capture that pasted material as part of your prompt, the agent may also read the contents of your clipboard, and on recent versions of macOS it relies on the system clipboard (pasteboard) permission to do so. This read is narrow by design: it happens only at the moment a paste into a supported AI tool is on screen, the clipboard text is used only when it verifiably matches that paste, and anything that does not match is discarded immediately. The agent does not otherwise monitor, read, or store your clipboard.

Data sharing

Kitestring shares the processed signals (classifications, counts, totals) with your organization, which deployed it. Kitestring does not sell your information or share it with third parties for their own purposes. Service providers acting on Kitestring’s behalf under contract (cloud hosting, AI model providers, email delivery, and the analytics provider described above) process data as subprocessors, bound to the commitments in this policy, or where disclosure is required by law.

Data retention

• Prompt text is retained while your account is active, so your coaching can build on how your work changes over time. Your organization can set a retention window, for example 90 days, after which prompt text is deleted. Short excerpts of your own prompts quoted in your personal coaching view are retained as part of your coaching history for as long as your account is active, and are visible only to you.
• Usage signals and classifications (counts, work-type and subject labels, token and cost estimates, response character counts, which AI tool was used, measured active time, and your organizational user identity) are retained for the duration of your organization’s subscription so dashboards and coaching can show trends over time, and are deleted or returned within 90 days after the subscription ends.
• Coaching guidance generated for you (the advice itself, not your prompt text) is retained while your account is active so your coach does not repeat itself.
• Aggregated or anonymized statistics that cannot be linked to an individual may be retained beyond these periods.

Your choices and rights

Kitestring is deployed by your organization, which decides whether and how it is used; your use may be governed by your organization’s workplace and monitoring policies. Because your organization is the controller of this data:

• Questions and requests about what is collected, how it is used, or to access or delete information associated with you should be directed to your organization first. Kitestring supports your organization in responding to these requests as its processor.
• Seeing your own data: you can view your own usage, classifications, and coaching in the Kitestring dashboard at any time.
• Legal rights: depending on your location you may have rights over your personal data (such as to access, correct, or delete it). Where these apply, they are exercised through your organization as controller, and Kitestring assists as required by law.

Contact

Questions about this policy can be directed to help@getkitestring.com.

Changes

We may update this policy. Material changes will be reflected by the “Last updated” date above.