How to use this file
What this is. New and changed wording for the privacy policy at https://proma.ai/privacy/ (hosted in Termly), so it covers Claude, ChatGPT and other AI apps that connect to Proma. There is also an optional addition to the terms of service at https://proma.ai/tos/.
This is not legal advice. Whoever owns legal for Growthzilla Inc. must read and approve this text before it is published.
When to publish. Draft it in Termly now. Publish only after the production release, because the text describes the new Proma approval page and Profile → Connected apps, which are not live yet.
How the part below the line works. Below the line, every line that starts with [WHERE: …] is for you. It says where the next block goes. Do not paste it. Paste only the text between one [WHERE] line and the next. Your final search for "[" will catch any you pasted by mistake. The [WHERE] lines also say which part each block belongs to:
- Part 1: a new section, "7. CONNECTED AI APPLICATIONS". Put it straight after section 6, "DO WE OFFER ARTIFICIAL INTELLIGENCE-BASED PRODUCTS?", and before "IS YOUR INFORMATION TRANSFERRED INTERNATIONALLY?". It sits next to section 6 because that section names Anthropic and OpenAI as Proma's own AI providers, and the new section explains that connecting Claude or ChatGPT is a different thing. Sections 7 to 17 move down by one. The policy refers to sections by name, not number, so nothing else breaks. Keep the heading wording: the other new lines refer to it by name.
- Part 2: small changes to existing sections, in the order they appear in the policy.
- Part 3 (optional): a new section for the terms of service.
In Termly. Add the text with Termly's option for your own wording (a custom section or clause). If you can't find it, ask Termly support. Make every "Learn more about connected AI applications" and every "Learn more in "CONNECTED AI APPLICATIONS"" a link to the new section, like the other "Learn more" lines in the summary. The new table of contents line, "7. CONNECTED AI APPLICATIONS", may turn into "1." when pasted, because editors read it as a numbered list. Type it by hand, or check it after pasting.
Dates. Set "Last updated" to the publish date on the privacy policy, and on the terms if you use Part 3. Termly may fill this in when you publish. Check that it did.
Placeholders to replace. When you are done, search the text you pasted for "[". Nothing should be left. If a link shows as "text" instead of a clickable link, turn it into a normal link in Termly.
- [Month DD, YYYY]: the publish date (one place, in Part 2).
- [N days]: how long logs of connector use are kept (two places: Part 1 and Part 2). Nobody has set this yet. An engineer checks the Cloudflare log settings (Workers logs and the log archive) and gives the real number, or sets one.
- [PENDING …] with [Version A] and [Version B]: one place in Part 1, about automations. Keep the version that matches the automation decision. Delete the other version, the PENDING line and the version label.
- [support@proma.ai]: this mailbox does not exist yet (one place in Part 1). Create it, or replace it with an address you already use.
Check before publishing
- Company name. The policy says "Growthzilla Inc." but the website footer says "Proma.ai, Inc.". Use one name everywhere. OpenAI checks this.
- Code emails. The text says Supabase checks passwords and codes and sends the code email. If the production Supabase project sends mail through a separate email service, name that service next to Supabase in Part 1 and in the "Service Providers" bullet in Part 2.
- The "sale or sharing" sentence added to section 13 in Part 2 is a legal judgement under US state privacy laws. Legal confirms it.
- Telling users. Section 15 says you may notify users of material changes. Legal decides whether this change needs an email.
- Numbers. The time limits in the text (about 10 minutes, about a minute, 1 to 15 minutes, 30 days, 90 days, 500 characters, about 120 actions a minute for each person in each organization) match the planned release. If engineering changes any of them before release, update the text.
- What the connector does today. Part 1 says a Connected App can see every automation in the organization and change or delete any of them, whatever the person's role, that it can create new spaces and systems, and that rows it adds are not marked with the person as their creator. That is how the connector works now. If engineering adds role checks to automations or records the creator before release, change those sentences to match.
- "Right away" in the app. The Connected apps page in Proma says an app loses access "right away". This text says "within about a minute", because a disconnect can take up to a minute to reach every server. Engineering may want the app wording to match.
Other problems in the current policy. The connector did not cause these, but fix them while you are editing, if legal agrees:
- "Cookie Notice: __________" is blank in three places (sections 1, 5 and 11).
- Section 4 names no service providers, although sections 1, 5 and 6 do. Part 2 offers a bullet that fixes this.
- Section 6 lists Anthropic, Google Cloud AI and OpenAI as AI Service Providers, but Proma's AI features also use Cloudflare Workers AI. Add Cloudflare, and check that "Google Cloud AI" is the right name for the Google service used.
- Section 13 says no personal information was disclosed to third parties for a business purpose, and marks category F, "Internet or other similar network activity", as NO. Both clash with Google Analytics advertising features (section 5), the AI providers (section 6) and usage logs (section 1).
- Section 15 says the new date is labelled "Revised", but the page labels it "Last updated".
- The table of contents calls section 3 "…YOUR PERSONAL INFORMATION?" but its heading says "…YOUR INFORMATION?".
- Termly's data for this policy shows
latest: false, which may mean a newer template version is waiting. Check in Termly before you edit.
<!-- ===== COPY EVERYTHING BELOW THIS LINE ===== -->
[WHERE: Part 1, new privacy policy section. Insert after section 6, "DO WE OFFER ARTIFICIAL INTELLIGENCE-BASED PRODUCTS?", and before "IS YOUR INFORMATION TRANSFERRED INTERNATIONALLY?"]
7. CONNECTED AI APPLICATIONS
In Short: You can let an AI application such as Claude or ChatGPT work in your Proma account. It acts as you, in one organization you choose, and only in systems you can open, except automations, which it can see across the organization. What it reads goes to that application's provider under their terms. You can disconnect it at any time.
You can connect third-party AI applications to your Proma account through our connector. Examples are Claude (made by Anthropic), ChatGPT and Codex (made by OpenAI), Cursor and similar tools. We call these "Connected Apps". You decide whether to connect each one. To connect an app, you sign in on a Proma page that shows the app's name and what it will be able to do. The app gets no access to your Proma information until you click Allow.
Knowing which app you are connecting. Proma recognizes a few well-known apps, such as Claude, ChatGPT and Grok, by the web address they send you back to, and shows their real name. For any other app, the name shown is the one the app gave itself. It could be any name, even "Claude", and we cannot check who runs it. If the app runs on your own computer, as Codex and Cursor usually do, the page says so and asks you to continue only if you just started it. Otherwise, the page warns you and shows the website you will be sent to. Only allow apps you trust.
What a Connected App can access
A Connected App acts as you:
- in one organization, which you choose when you connect it. It cannot reach your other organizations.
- only in the systems your Proma role lets you open. Automations are the exception: it can see every automation in the organization, even in systems you cannot open.
- with the access the app asks for and you approve: to see information, or to see and change it. The approval page lists what the app asks for. You either allow all of it or click Cancel.
With access to see information, it can read:
- the content in those systems, such as datasets and their rows, and forms and their settings. This can include personal information about your customers, contacts or colleagues, such as names, email addresses and phone numbers.
- every row of the datasets in those systems, even rows that your usual views or interfaces filter out.
- details about people in your organization where they appear in that content, for example in a Team Members or approval column: their name, email address, phone number, role and department, and who approved a request. For each record it can also see the ID of the member who created it, and when it was created and last changed.
- how your workspace is set up: the names of spaces, systems, datasets, columns, roles and interfaces.
- every automation in your organization and its run history, including the data each step received and produced.
- the names of the outside accounts your automations use, such as a Slack or Gmail connection, and lists of choices it fetches live from those accounts while setting up an automation, such as channel, file or contact names. It cannot read their passwords or keys.
With access to change information, it can also add and update rows, build and change systems, forms and interfaces, and delete columns, interfaces and datasets. Many apps ask you to confirm before deleting, but Proma cannot check that they do. Deleting a dataset also deletes its rows. It can change data and structure only where you are an admin, either of that system or of the organization. It can also create new spaces and systems.
[PENDING — use Version A if every Connected App can build automations, Version B if ChatGPT can only view automations. Delete the other version, this line and the version label.]
[Version A] With access to change information, it can also create and publish automations, and switch on or off or delete any automation in the organization, whatever your role. A published automation runs by itself. It can send information from your organization to other services and web addresses, either through connections your organization has set up in Proma, such as email or chat tools, or through a web request or code step that can send data to any address.
[Version B] With access to change information, most Connected Apps can also create and publish automations, and switch on or off or delete any automation in the organization, whatever your role. A published automation runs by itself. It can send information from your organization to other services and web addresses, either through connections your organization has set up in Proma, such as email or chat tools, or through a web request or code step that can send data to any address. ChatGPT can only view your automations and their run history. It cannot create or change them.
Changes an app makes are saved to your data like other changes. Rows an app adds are not marked with you as their creator.
Where your information goes. When a Connected App reads information from Proma, we send that information to the app. From then on, the app and its provider hold it under their own terms and privacy policy, not ours. Proma does not control how they use, store or share it. Please read their terms before you connect.
Anthropic and OpenAI are also named in "DO WE OFFER ARTIFICIAL INTELLIGENCE-BASED PRODUCTS?" as providers of Proma's own AI features. When you connect Claude or ChatGPT yourself, that is a separate arrangement between you and that company, and our agreements with them do not cover it. Proma does not use its own AI to handle a Connected App's requests; the app's own AI does that work. If information an app adds reaches a Proma AI feature, such as an AI column or an automation with an AI step, that section applies to it as usual.
Signing in to connect an app. You sign in with your Proma email address and either your password or a 6-digit code we email you. Our authentication provider, Supabase, checks your password or code and sends the code email. Connecting an app does not create a Proma account.
What we keep about each app. The details the app registered with us, such as its name, its web addresses and any contact details its developer gave. We delete these after 90 days without use, even if you disconnect sooner.
What we keep for each connection. The organization and the access you approved, when you connected it, and when it will expire if it is not used.
The keys that give an app access are stored only in a scrambled form that cannot be turned back into the key. The connection is filed under your Proma user ID, with the access you approved. The organization and sign-in details stored with it are encrypted.
What we keep for a short time
- A sign-in record. While you connect an app, we hold your email address, the app's details, the access it asked for and the organizations you can choose from. We delete it when you click Allow or Cancel, or after about 10 minutes.
- Abuse-prevention counters. To stop abuse, we count sign-in attempts, code emails and connection requests for each network (IP) address and email address, and the actions each person's apps run. These counters are deleted after 1 to 15 minutes. Most of them hold the address only in a scrambled form.
Logs. We keep logs of connector use to keep the service secure and working, and to investigate problems. They record events such as:
- when the approval page is shown, and whether you clicked Allow or Cancel
- failed sign-in attempts, and when a limit was reached
- the actions an app runs. For reading and writing data, we log your account, the organization, which system and dataset it used, and how many rows it read or wrote. For building tools, we log how long the action took and whether it worked.
- when you disconnect an app
Logs record which systems and datasets an app used and how many rows, not what is in them, with two exceptions: the wording of a query an app runs to answer a question about your data (the first 500 characters), and error messages. Both can include values from your data, such as a name the app searched for. We do not log your password, your sign-in codes or the keys that give an app access. Logs can also include the technical details described under "Log and Usage Data" in "WHAT INFORMATION DO WE COLLECT?". We keep these logs for [N days] and then delete them.
Your choices
- See your Connected Apps in Proma under Profile → Connected apps. The list shows the apps you connected in the organization you are signed in to. To see apps in another organization, switch to it first.
- Click Disconnect next to an app. If you have more than one, you can also click Disconnect all to disconnect every app in this organization. The app loses access within about a minute, and we delete the connection record and its access keys. An action it had already started may still finish.
- You can also remove Proma inside the AI app, but we cannot see or check that. The connection stays active on our side until it goes 30 days without use. To be sure access has ended, disconnect the app in Proma.
- A connection that is not used for 30 days expires and is deleted automatically.
- Disconnecting does not delete information the app already received. To ask for that, contact the app's provider.
Questions. Email privacy@proma.ai. For help connecting an app, see proma.ai/docs/mcp or email [support@proma.ai].
[WHERE: Part 2, changes to existing sections. At the top of the policy, replace "Last updated March 04, 2025" with:]
Last updated [Month DD, YYYY]
[WHERE: Part 2. In the opening list (the one that starts "Visit our website at https://proma.ai"), after the bullet that starts "Use Proma.ai.", add this bullet:]
Connect a third-party AI application, such as Claude or ChatGPT, to your Proma account
[WHERE: Part 2. In "SUMMARY OF KEY POINTS", replace the whole paragraph that starts "Do we collect any information from third parties?" with:]
Do we collect any information from third parties? Apart from what an AI application you connect sends us, meaning the details it registers with us and the requests it makes on your behalf, we do not collect any information from third parties. Learn more about connected AI applications.
[WHERE: Part 2. In "SUMMARY OF KEY POINTS", after the paragraph that starts "In what situations and with which parties do we share personal information?", add:]
Can I connect AI applications such as Claude or ChatGPT? Yes. A connected application acts as you, in one organization you choose, and only in systems you can open, except automations, which it can see across the organization. Information it reads goes to that application's provider under their own terms. You can disconnect it at any time. Learn more about connected AI applications.
[WHERE: Part 2. In the "TABLE OF CONTENTS", after "6. DO WE OFFER ARTIFICIAL INTELLIGENCE-BASED PRODUCTS?", add the line below. Type it by hand if your editor turns it into "1.". Then renumber the lines after it, so 7 becomes 8 and so on, up to 17 becoming 18:]
CONNECTED AI APPLICATIONS
[WHERE: Part 2. In section 1, "WHAT INFORMATION DO WE COLLECT?", after the paragraph that starts "Payment Data.", add:]
Information from connected AI applications. If you connect a third-party AI application to your account, we receive the details it registers with us, such as its name, its web addresses and any contact details its developer gives. We also receive the requests it sends on your behalf, such as a question about your data or rows to add. Learn more in "CONNECTED AI APPLICATIONS".
[WHERE: Part 2. In section 2, "HOW DO WE PROCESS YOUR INFORMATION?", after the bullet that starts "To deliver and facilitate delivery of services to the user.", add this bullet:]
To connect AI applications you choose. We may process your information to confirm who you are when you connect an AI application, check what it is allowed to do, carry out the requests it sends on your behalf, and protect the connection from abuse.
[WHERE: Part 2. In section 4, "WHEN AND WITH WHOM DO WE SHARE YOUR PERSONAL INFORMATION?", after the bullet that starts "Business Transfers.", add this bullet:]
Connected AI Applications. When you connect a third-party AI application to your account, we share with it the information it requests on your behalf, within the access you approved. The application's provider handles that information under its own terms. Learn more in "CONNECTED AI APPLICATIONS".
[WHERE: Part 2, recommended. In the same section 4, also add this bullet:]
Service Providers. We share information with companies that provide services to us under contract. These include hosting (Amazon Web Services, Cloudflare and Supabase), sign-in and sign-in emails (Supabase), payments (Stripe, Wise and PayPal), analytics (Google Analytics) and the AI features described in "DO WE OFFER ARTIFICIAL INTELLIGENCE-BASED PRODUCTS?".
[WHERE: Part 2. In section 6, "DO WE OFFER ARTIFICIAL INTELLIGENCE-BASED PRODUCTS?", after the last paragraph (the one under "How We Process Your Data Using AI"), add:]
AI applications you connect yourself. Connecting Claude, ChatGPT or another AI application to your Proma account is different from the AI Products described in this section. When you use an application this way, its provider is not acting as our AI Service Provider. Learn more in "CONNECTED AI APPLICATIONS".
[WHERE: Part 2. In section 8, "HOW LONG DO WE KEEP YOUR INFORMATION?", after the paragraph that ends "past the termination of the user's account.", add:]
Connected AI applications. We keep the record of a connection until you disconnect it or it goes 30 days without use. An app's registration details are deleted after 90 days without use. Sign-in records for connecting an app are deleted after about 10 minutes, and abuse-prevention counters after 1 to 15 minutes. Logs of connector use are kept for [N days]. Information a connected application has already received is kept by its provider under its own terms.
[WHERE: Part 2. In section 11, "WHAT ARE YOUR PRIVACY RIGHTS?", after the paragraph that starts "Upon your request to terminate your account", add:]
Connected AI applications: You can see and disconnect the AI applications connected to your account at any time in Proma, under Profile → Connected apps, one organization at a time. Disconnecting ends the application's access within about a minute. It does not delete information the application already received. To ask for that, contact the application's provider.
[WHERE: Part 2. In section 13, "DO UNITED STATES RESIDENTS HAVE SPECIFIC PRIVACY RIGHTS?", after the sentence that starts "We have not disclosed, sold, or shared any personal information", add:]
This does not include information you tell us to send to an AI application you connect to your account. We send that information at your direction, and we do not treat it as a sale or sharing of your personal information.
[WHERE: Part 3, optional, terms of service. After section 10, "THIRD-PARTY WEBSITES AND CONTENT", add the section below. Then add it to the table of contents, renumber the later sections, and set "Last updated" to the publish date.]
11. CONNECTED AI APPLICATIONS
You may connect third-party AI applications, such as Claude or ChatGPT, to your account through our connector ("Connected Apps"). Using a Connected App this way, as described at proma.ai/docs/mcp, is an approved use of the Services. It does not by itself break clause (5) of "USER REPRESENTATIONS" or the rules on automated use in "PROHIBITED ACTIVITIES".
When you connect an app:
- You choose which apps to connect and what access to give them. Only connect apps you trust.
- You are responsible for what you ask a Connected App to do, and for every action it takes through a connection you approved, as if you had taken it yourself. This includes changes and deletions of data, and any automations it publishes for you.
- AI applications can make mistakes. Check what an app plans to change before you confirm it, especially a deletion.
- A Connected App is a third-party service under its own terms, and "THIRD-PARTY WEBSITES AND CONTENT" applies to it. We are not responsible for a Connected App's actions or outputs, including content it creates, mistakes it makes, or what it or its provider does with information it receives.
- You must have the right to share any information you let a Connected App read, and you must follow that app's terms.
- We may limit how often a connection can be used, currently about 120 actions a minute for each person in each organization, and we may change these limits. We may suspend or disconnect a Connected App, or limit your use of the connector, if we see abuse, a security risk, or use that breaks these Legal Terms.