everyais studio Privacy Policy

1. Scope

BROZ Inc. (the "Company," "we," "us," or "our") publishes this Privacy Policy to protect personal information and respond promptly to privacy concerns under the Personal Information Protection Act of Korea and other applicable law. It applies to everyais studio (the "Service") at studio.everyais.com.

2. Why We Process Information

• confirm registration, authenticate through Firebase, and manage accounts and consent records;
• generate and edit images and videos from prompts and settings and provide job progress, outputs, and history;
• issue, verify, and revoke API Keys, approve, verify, and disconnect Connected Apps (External Agents connected by OAuth), process External Agent requests, and manage per-key and per-connection usage caps and per-account request limits, if you use that feature;
• provide the conversational everyagents (interpreting your requests, suggesting models and settings, presenting server quotes, starting the generations and everyflow runs you accept or that fall within your auto-start limit, creating and editing everyflows, researching by web search and by reading web pages and images, saving your skills and memory, and summarizing long conversations), charge Credits for each reply, and manage everyagents usage limits, if you use that feature;
• provide everyflow and everyflow apps (storing documents, running nodes and keeping their results, and publishing, sharing by link, and running apps), run AI text nodes, charge Credits for each run, and manage their usage limits, if you use that feature;
• manage Credits, payments, subscriptions, coupons, generated-data retention, and transaction records;
• review and publish eligible free-credit outputs in the gallery and process consent withdrawals;
• handle business inquiries, feature and model requests, support, and service notices;
• measure usage and advertising performance and improve and market the Service where consent or another lawful basis applies;
• detect and prevent abuse, duplicate benefits, security threats, and incidents; and
• comply with legal obligations and handle disputes.

3. Information We Process and How We Collect It

We collect information you enter on registration, profile, generation, checkout, and inquiry screens and in tickets in our Discord support server; information created automatically when you use the Service; and information received from connected providers such as Firebase and our payment processor. We do not directly store your full card number or password.

4. Retention

Where the law requires continued retention, we segregate the record and use it only for that purpose. User-generated data is excluded from cross-account backup; after access is blocked at expiry, every object version and delete marker is permanently removed by the next deletion job.

5. Disclosure to Third Parties

We do not disclose personal information to third parties except with your consent or as permitted by law. The following disclosures occur when necessary to complete a generation or payment you request, to reply to you in everyagents, or to run an everyflow AI text node.

Each AI Provider receives information only when you request a generation with its model. The same applies to generations requested through an External Agent with an API Key or a Connected App, to generations started in everyagents by accepting a quote or within the auto-start limit, and to generations made by the nodes of an everyflow or an everyflow app. Google LLC, however, also receives conversation content (including attached images, web pages and images read, and memory and skills) to generate everyagents's replies, even when no generation is requested, and receives search queries when a Parallel web search fails and the web is searched through Grounding with Google Search (Section 9-3). When you run an everyflow AI text node, its inputs go to the AI Provider of the model chosen in that node (Section 9-4).

The payment method, billing address, and other details you enter on the USD checkout are collected directly by that order's Merchant of Record, Paddle or Dodo Payments, and are governed by its privacy policy (Paddle: https://www.paddle.com/legal/privacy; Dodo Payments: https://dodopayments.com/privacy-policy).

6. Service Providers

We use contractual and operational controls to require service providers to protect personal information and disclose changes to providers or processing through this Policy.

7. International Transfers

You may object by stopping the relevant feature or requesting account deletion. Essential transfers are required for features such as authentication and image generation, so refusing them means those features cannot be provided. If you email portal@everyais.com for support instead of using Discord, nothing is transferred to Discord Inc. or Ticket Tool.

8. Cookies, Analytics, and Advertising

• A language cookie remembers your choice for up to one year.
• A first-touch cookie measures campaign and signup attribution for up to 90 days.
• Firebase may use storage or tokens necessary for authentication state and security.
• Depending on production configuration, Google Analytics, Google Ads, Meta Pixel, and X Pixel may process page-view, product-usage (such as generation and download), registration, purchase, and related events.
• For signed-in members, these events carry a member identifier and a SHA-256 hash of the email address (and of the mobile number for payments) for advertising measurement and conversion analysis. The unhashed email address and mobile number are not sent to advertising platforms.
• You can delete or block cookies in your browser and limit personalized advertising in the relevant platform settings. Blocking essential cookies may prevent sign-in and other features from working.

9. Optional Gallery Consent

Gallery publication is optional and separate from processing required to provide the Service. If an opted-in User generates an image entirely with free Credits, we may review it and publish the approved image together with its prompt, model, and resolution.

We do not publish an image paid for in whole or part with purchased or subscription Credits. You may withdraw consent at any time in Settings, which removes published and pending entries. Declining or withdrawing does not affect core image-generation access.

9-2. API Keys, Connected Apps, and External Agents

This Section applies if you issue an API Key in Settings, or approve an External Agent's OAuth connection request (a Connected App), and connect an External Agent you choose, such as an AI tool, to the Service's MCP endpoint (everyagents MCP). Retention periods are set out in Section 4.

• Key information: we store the key name, a key prefix for telling keys apart, the SHA-256 hash of the key, the 24-hour usage cap, the issuance, expiry, revocation, and last-use times, and the version and time of the usage conditions you accepted when issuing the key. The full key is shown once at issuance and is not stored.
• Connection information: when you approve a connection, we store for each connection your member identifier, the External Agent's name and redirect addresses (as the agent registered them), the 24-hour usage cap, the version and time of the API usage conditions you accepted, and the connection, expiry (365 days after approval), disconnection, and last-use times. Each approval writes a server log line with your member identifier, the External Agent's identifier, and the redirect address's host.
• Authorization codes and tokens: authorization codes (valid for 10 minutes, usable once), access tokens (1 hour), and refresh tokens (90 days, replaced on each use) are given in full only to the External Agent; instead of the full values, our servers keep their SHA-256 hashes with your member identifier, the External Agent or connection each belongs to, and their issue, expiry, and use times. An authorization code also keeps the approval request (the redirect address, the PKCE challenge, the scope and target endpoint, the 24-hour usage cap, and the version and time of the API usage conditions you accepted). If a refresh token that was already used is presented again, the connection is ended at once.
• Generation records: a generation job requested through an External Agent also records the request channel and the identifier of the key or connection used, and an everyflow or app run started through an External Agent records that identifier too, which counts the run toward the usage cap. Prompts, inputs, and outputs are stored in generation history, as with generations requested on the web.
• Direct uploads: on an External Agent's request we issue an upload URL usable for 15 minutes, and the image uploaded to it (PNG, JPEG, or WebP up to 15MB) is checked, converted, and stored as a reference image when it is first used. It is kept like any other reference image.
• Request records: each External Agent call writes a server log line with the key or connection identifier, the feature called, the job or run identifier, the outcome, the Credits reserved, quoted, or returned by the call, and the processing time. These records do not include prompts or the full key, nor any full token.
• Server request records: Google Cloud (Cloud Run), which runs our servers, automatically creates a request record for each request, including the request time, IP address, client information, request URL, and response status, and keeps it in Google Cloud Logging.
• Model lists, quotes, job status and output links, lists of your everyflows and apps (names, descriptions, and input and output settings), the status and results of everyflow and app runs (including text written by AI text nodes), and direct-upload URLs sent to an External Agent are provided to you at your request. The External Agent's provider is not the Company's processor or a third-party recipient, and its processing is governed by the terms between you and that provider.

All API Keys and Connected Apps (authorization codes and tokens included) are deleted when you delete your account. You can revoke a key or disconnect an app in Settings at any time, and if the Company forcibly ends a member's sessions, all of that member's API Keys and connections are revoked as well.

9-3. everyagents

This Section applies if you use the Service's everyagents, a chat agent that interprets your requests in conversation, suggests models and settings, shows quotes calculated by our servers, creates or edits everyflows on request, and finds material by searching the web and reading web pages and images. Retention periods are set out in Section 4.

• Stored information: for each conversation we store its title (the first line of your first message), the messages you send, everyagents's replies, everyagents's function calls and their results (model identifiers, generation settings, prompts, plans, and everyflow documents everyagents wrote, job, output, and everyflow identifiers, web search results (deleted when that reply ends), the addresses or identifiers of web pages and images read, the text of web pages read, the research given to explorer sub-agents, explorer sub-agent reports, and memory content read or written), web search records (search queries, and the titles and addresses of the sources found; the summary of what was found and Google Search suggestions are deleted when the reply ends), conversation summaries, each quote card and its outcome (the Credits shown, the settings you changed on the card and its requotes, and whether and when it was accepted, declined, or closed by a new message), each skill save card and its outcome (the proposed skill's name, description, and body, and whether and when it was saved or declined), the outcome of each proposed everyflow change (whether and when it was approved, rejected, or closed by a new message), automatic-start records (the Credits started and when), and the conversation's settings (Plan mode and processing level, auto-start limit, and the everyflow it belongs to). You can attach up to 3 images to a message (newly uploaded or picked from your outputs); the conversation keeps only the identifiers of attached images and chosen skills, and stores no image files or output links. When you copy a conversation with 'New conversation from here', its record up to that message (function calls and results, web search records, and conversation summaries included) is copied into the new conversation.
• Information sent to Google: when you send a message, run a plan, decline a quote, save or decline a skill on its save card, answer a question, approve or reject a proposed everyflow change, or continue after a generation finishes, everyagents sends Google LLC's Gemini API the recent part of that conversation's messages, replies, and function calls and results, including the settings you changed on a quote card, web search results, the text of web pages read, explorer sub-agent reports, and memory content read or written (prior conversation context), a summary of the part before it (conversation summaries), the Company's instructions and function descriptions, the descriptions in the skill list and the instructions of the skills everyagents reads or you choose (including skills you saved), the list of memory files (path, size, and modification time), and the latest 6 images attached to the conversation (reduced to 1024px JPEG) and the images everyagents views while researching, to generate a reply. Web search results are sent only while generating the reply that received them, never with later replies. These web search results include those Parallel finds. A search query is sent to Grounding with Google Search for searching only when the Parallel web search fails (see Web search). Videos are not sent, and neither is account information such as your name or email. Accepting or requoting a quote does not by itself send conversation content; the accepted generation is shared with the AI Provider of the chosen model under Section 5.
• Request records: each time everyagents handles a request, we write a server log line with the conversation identifier, the request type, the number of steps, the result code, the AI Provider's response status, the AI cost in USD and the Credits charged, the number of quote items and their Credits, the number of jobs started, the Credits started automatically, the processing time, and the processing level and model used for the reply. These records do not include the content of messages or replies.
• AI usage records: each Gemini API or Parallel call by everyagents (web search, explorer sub-agent, and conversation-summary calls included) records your member identifier, the model, the outcome, token counts, the cost in USD (web search fees included) and whether it was estimated, the price-table version applied, the processing time, and the timestamp. These records contain no conversation content or conversation identifier and are used only to calculate the Credits charged for each reply and to manage the Service's overall usage limit.
• Generations: everyagents starts a generation or an everyflow run only when you review and accept a quote card calculated by our servers, or when the quoted total is within the auto-start limit you set for the conversation (0 by default, at most 2,000 Credits per request) and enough Credits are available; otherwise it reserves or spends no generation Credits. In Plan mode it starts nothing until you run the plan, and after that only in the cases above. A generation or everyflow run started this way also records the request channel and the conversation identifier, and its prompts, inputs, and outputs are stored in generation history, as with generations requested on the web.
• Research: on request, everyagents reads web page addresses from our servers and uses their text, and views your uploaded or output images or external images. When it reads a web page, our servers request that address and send no account information about you. It views only a copy of an external image that our servers downloaded, and that copy is deleted after 24 hours like other reference images (Section 4). It cannot view other Users' outputs. everyagents may hand research to an explorer sub-agent, which works within the same request, reads only web pages through the Gemini API (it views no images and does not search the web), and returns only a report; it does not see this conversation and receives only the research everyagents gave it. The explorer sub-agent's own conversation is not stored; its report is stored as a function result of this conversation.
• Web search: everyagents first sends search queries to Parallel Web Systems, Inc. (United States) to search the web and images. For web search, it sends a search query to Google LLC's Gemini API, which uses Grounding with Google Search, only when the Parallel search fails. Image search runs on Parallel only and is never replaced by Grounding with Google Search when the Parallel search fails. everyagents writes search queries from the conversation, so they may include what you wrote there; these search calls never send the conversation. Parallel receives only the search query, and a Grounding with Google Search call receives only the search query and the Company's instructions. A web search first calls Parallel's search under the Company's Parallel account. Parallel keeps the search queries it receives, and their results, as provided by its policy and currently sets no retention period, keeping them for support and debugging (Section 7). Google stores the prompts, context, and output of grounding requests for 30 days to create grounded results and Google Search suggestions and to debug and test the systems that support Grounding with Google Search; for paid use, that debugging and testing follows Google's data processing addendum. The conversation stores the search query and the titles and addresses of the sources found. The summary of what was found and the Google Search suggestions are kept in the conversation only until that reply ends and are then deleted, and the Google Search suggestions are shown on the screen of the reply that used the search. Search results are sent to the Gemini API only to generate the reply that made the search, never with later replies, so a follow-up question that needs them searches again and incurs the web search fee again. Search results are shown only to the User of that conversation and are never shared with other Users or cached for them. Web search is not offered to External Agents (everyagents MCP).
• Conversation summaries: when what a conversation would send to the Gemini API exceeds 75% of its limit, everyagents summarizes the oldest part through the same Gemini API, stores the summary in the conversation, and sends it from then on instead of that part. The summary is not shown on screen and is kept as long as the conversation. The AI processing cost of the summary is part of that reply's charge.
• Memory: everyagents creates and edits memory files (path, content, and modification time) on your account and uses them in all your conversations. Each time it writes a reply it sends the Gemini API the list of memory files (path, size, and modification time), and it sends a file's content as that conversation's function calls and results when it reads or writes the file. Memory files are kept until you delete them or your account. You can view and delete memory files, one or all, but not edit them. Memory content read or written is also stored as that conversation's function results and stays for the conversation's retention period; deleting a memory file does not delete it from that conversation's record. Memory is not offered to External Agents (everyagents MCP).
• Your skills: we store on your account the skills (name, description, and body) you approve on a save confirmation card, up to 20 per account, each with a body of up to 8,000 characters; saving again under the same name replaces that skill. The names and descriptions of saved skills are sent to the Gemini API as the skill list each time a reply is written, and a body is sent when everyagents reads it or you choose it. Your skills are kept until you delete them or your account, and you can view and delete them. Your skills are not offered to other Users or to External Agents (everyagents MCP).
• Reply charges: when everyagents produces a reply, we deduct Credits equal to 1.5 times the AI processing cost (in USD, including explorer sub-agent and conversation-summary calls and web search fees) of that reply after it finishes (Article 4 of the Terms and the Credit Pricing Policy), and record the conversation identifier, the AI cost in USD, and the number of calls in your Credit history. Web search fees are calculated at the AI Provider's published price per query, without deducting any free allowance the provider offers. Calls whose cost was recorded as an estimate because the AI Provider reported no usage after an error or interruption are not charged.
• AI replies: everyagents's replies and suggestions are AI-generated and may be wrong or inaccurate, and the Service says so on screen.
• Google's processing: the Company uses the Gemini API as a paid service. Google does not use prompts and responses from paid use to improve its products. Google logs requests from paid use for a limited period only to detect and prevent prohibited use, keep the services safe and secure, and make disclosures required by law. everyagents does not use features that keep conversations on Google's side (the Interactions API's storage option). In addition, for the Grounding with Google Search requests used for web search, Google stores the prompts, context, and output for 30 days to create grounded results and Google Search suggestions and to debug and test the systems that support Grounding with Google Search (for paid use, under Google's data processing addendum).

All everyagents conversations, memory files, and your skills are deleted when you delete your account. AI usage records, which contain no conversation content, remain linked to the anonymized account until they are deleted 3 months (90 days) after the call, and per-reply Credit charge records are kept like other Credit records under the payment-record retention in Section 4.

9-4. everyflow and everyflow Apps

This Section applies if you use everyflow, which connects nodes to chain generation and AI text steps, and everyflow apps. Retention periods are set out in Section 4.

• everyflow documents: for each everyflow we store its name and document (its nodes and connections, each node's prompts and settings, and the upload and output identifiers that image nodes point to).
• Node results: each time all runs of a node finish, we store its result (an AI text node's text, or output identifiers) and a fingerprint (hash) of its settings and inputs, keeping only the latest 20 for each node.
• Run records: for each run we store the document as it was run, the run's scope, each node's run count and quote, the results reused from nodes above, the text written by AI text nodes, and the status and failure reason. Generation jobs a run makes also record the request channel and the run identifier, and their prompts, inputs, and outputs are stored in generation history, as with generations requested on the web.
• AI text nodes: when a node runs, its instructions and the text (at most 20,000 characters per run) and images (at most 4, reduced to 1024px JPEG) of the nodes connected to it are sent to the AI Provider of the model chosen in the node (Google LLC, OpenAI OpCo, LLC, or Anthropic, PBC). The default model is Google Gemini, and the models a node can use are those the Service offers in the node among Google Gemini, OpenAI gpt-6-astra and gpt-6.1-sol, and Anthropic Claude Opus 5.5. Account information such as your name or email is not sent; each provider's processing is as described in Section 5, and Google's as described in Section 9-3. Each call records an AI usage record with the same items as in Section 9-3, identified by a charge key made of the run identifier, the node key, and the run's index. AI usage records are used to calculate the Credits charged for each run and to manage the Service-wide and per-account usage limits.
• AI text node charges: each run is charged 1.5 times its AI processing cost in USD (at most the maximum Credits of the model chosen in the node; each model's maximum is shown on the model selection screen and in quotes, and the default model's is 63 Credits) in Credits (Article 4-2 of the Terms and the Credit Pricing Policy), and the run identifier, node key, run index, and AI cost in USD are recorded in your Credit history. Calls whose cost was recorded as an estimate because no usage was reported are not charged.
• Publishing an app: we store the app's name, description, sharing setting, and link token, and with each publish a copy of the document (a version) and its input and output settings (labels, descriptions, and whether required). A person who opens the app by link sees its name and description, its inputs' labels, descriptions, and whether each is required, and its outputs' labels, and also receives its input nodes' keys, never its document or prompts.
• Running an app: an app run's record is stored on the running User's account and holds a copy of the app's document as run (including the publishing User's step prompts) and the running User's inputs. The running User is shown only the output nodes' results, but each step's generation job and prompt remain in the running User's generation history. The publishing User cannot see other Users' run records.
• Deleting an app: when the publishing User deletes an app, it is hidden and its link stops working. Its versions stay because other Users' run records point to them.
• everyagents and External Agents: on your request, everyagents may create or edit your everyflows and run your everyflows and apps (Section 9-3), and an External Agent may run your everyflows and the apps you can open (Section 9-2). What they do is stored as described in this Section.

When you delete your account, we delete your everyflow documents and node results, your own run records, and the records of apps you ran. Apps you published are hidden and their links stop working; each version's document, inputs, and output labels are erased, keeping only the output nodes' keys and types that other Users' run records use to show their results. Records of other Users who already ran your apps (the copy of the app document and the generation history on their accounts) remain as their information. AI text node Credit charge records are kept like other Credit records under the payment-record retention in Section 4.

10. Your Rights

• request access, portability, correction, deletion, or restriction and withdraw consent;
• withdraw marketing-email and gallery-publication consent in Settings;
• delete individual everyagents conversations in the product (deleting a conversation does not delete the generation jobs and outputs it started; they follow the generation-record retention in Section 4);
• view and delete everyagents memory files and your skills in the product (deleting a memory file leaves the record of conversations that read or wrote it until those conversations are deleted);
• delete everyflows and everyflow apps in the product (deleting an app leaves other Users' run records on their accounts) and disconnect Connected Apps in Settings; and
• exercise rights through a lawful representative.

Submit a request through Settings or to portal@everyais.com. After verifying you or your authorized representative, we respond within the period required by law. A request may be limited where retention is required by law or it would infringe another person's rights.

11. Deletion and Security

We delete personal information without undue delay when its retention period expires or its purpose is complete. Electronic files are deleted using methods designed to prevent recovery, paper is shredded or incinerated, and records required by law are segregated.

Our safeguards include encryption in transit, least-privilege access, database access controls, encryption and key management for payment keys, abnormal-request controls, audit logs, private image storage, and expiring signed delivery URLs. API Keys are stored only as SHA-256 hashes and shown in full once at issuance; every key has an expiry date and a 24-hour usage cap, and you can revoke a key in Settings at any time. OAuth authorization codes and tokens are likewise stored only as SHA-256 hashes; every connection has an expiry date and a 24-hour usage cap, and you can disconnect an app in Settings at any time.

12. Privacy Contact and Remedies

• Privacy officer: Seungyun Kim
• Title: Representative director
• Email: nilk@broz.co.kr
• Phone: +82-10-2798-3438
• Company: BROZ Inc.
• Business registration no.: 288-81-02136
• Address: Room 202-M10, 2F, Bldg 1, 5 Hyosin-ro 16-gil, Dong-gu, Daegu, Republic of Korea

In Korea, you may also contact the Personal Information Dispute Mediation Committee (1833-6972), the Privacy Infringement Report Center (118), the Supreme Prosecutors' Office (1301), or the National Police Agency (182).

13. Changes to This Policy

We announce additions, deletions, or amendments at least 7 days before they take effect. We provide at least 30 days' notice of a change that materially affects User rights and obtain renewed consent where required.

This Policy takes effect on October 27, 2026, except that the amendments of October 18, 2026 take effect on October 18, 2026. Amended on October 27, 2026: USD payments are sold by Paddle or Dodo Payments as Merchant of Record, so the Dodo Payments Inc. entries in Sections 5 and 7 now cover all USD payments rather than only the orders of October 23 to 25, 2026. Amended on October 26, 2026: Dodo Payments Inc. (United States) still handles the sale and refunds of the USD orders paid through Dodo Payments from October 23 to October 25, 2026, so it is listed again, for those orders only, as a third-party recipient (Section 5) and an overseas recipient (Section 7). Amended on October 25, 2026: the seller of USD payments changed from Dodo Payments Inc. to our reseller Paddle, so Dodo Payments Inc. (United States) was removed and Paddle.com Market Limited (United Kingdom) and Paddle.com Inc. (United States) were added as third-party recipients (Section 5) and overseas recipients (Section 7). Amended on October 23, 2026: USD payments are offered again through Dodo Payments Inc. (United States) as Merchant of Record, so Dodo Payments Inc., which receives your member identifier, email address and name, and order information when you start a USD checkout, was added as a third-party recipient (Section 5) and an overseas recipient (Section 7). Amended on October 19, 2026: Meta Platforms, Inc. (United States), which receives information when you request a generation with Meta Muse Image 1.0, and Ideogram AI, Inc., which receives information when you request a generation with an Ideogram model, were added as third-party recipients (Section 5) and overseas recipients (Section 7). Amended on October 18, 2026: the items (Section 3) and retention period (Section 4) of customer support requests received as tickets in our Discord support server were added, and Discord Inc. (United States) and Ticket Tool were added as service providers (Section 6) and overseas recipients (Section 7). Amended on October 16, 2026: USD payments are no longer offered, so Polar Software Inc. (United States), the former seller of USD payments, was removed from the third-party recipients in Section 5 and the overseas recipients in Section 7. Amended on October 15, 2026: Sections 7 and 9-3 now state that the Parallel web search no longer sends a per-conversation identifier and is called with the Company's Parallel account. Amended on October 15, 2026: everyagents's web search (Section 9-3) now sends search queries to Parallel Web Systems, Inc. (United States) first, searches the web through Grounding with Google Search only when the Parallel search fails, and runs image search on Parallel only. Parallel Web Systems, Inc. was added as a service provider (Section 6) and an overseas recipient (Section 7), Google's web search from search queries in Sections 5, 6, and 7 is now limited to when a Parallel web search fails, and AI usage records now include Parallel calls. Sections 7 and 9-3 now state that a web search goes first to Parallel's free search and that its call also carries a per-conversation identifier (derived from the conversation identifier so that it cannot be turned back into it). Amended on October 15, 2026: everyagents (Section 9-3) now covers the items and retention of research (reading web pages and images, and explorer sub-agents), web search through Grounding with Google Search with its 30-day storage and the Google Search suggestions shown, conversation summaries, memory files, and your skills, and the information sent to Google now includes the skill list's descriptions, the list of memory files, web pages and images read, and conversation summaries. Reply charges now state that they include explorer sub-agent and conversation-summary calls and web search fees. Section 5 now states that Google stores requests that use Grounding with Google Search, including search-grounded image generation, for 30 days. An everyflow AI text node (Section 9-4) now sends its inputs to the AI Provider of the model chosen in the node, so the items disclosed to OpenAI OpCo, LLC were widened and Anthropic, PBC was added as a recipient (Section 5) and an overseas recipient (Section 7), and the maximum charge per run now depends on the model. The retention of memory files and your skills (Section 4) and the right to view and delete them (Section 10) were added. Amended on October 14, 2026: everyagents (Section 9-3) now states that Plan mode starts no generation until you run the plan, and after that only in the same cases. The plans everyagents writes, the conversation's processing level, and the settings you changed on a quote card and its requotes were added to the stored items in Sections 3 and 9-3, and the processing level and model used for each reply were added to request records. Conversation content is also sent to Google LLC when you run a plan, and it includes the settings you changed on a quote card. Amended on October 12, 2026: the purposes, items, and retention of everyflow and everyflow apps (Section 9-4) were added. Connected Apps (OAuth connections) and External Agents' direct uploads were added to Section 9-2, and their retention and that of reference images to Section 4. Sections 5, 6, and 7 now state that everyflow AI text node inputs and images attached in everyagents are disclosed to, processed by, and transferred overseas to Google LLC. everyagents (Section 9-3) now covers attached images, skills, conversation settings, creating everyflows, and starting within the auto-start limit. How Connected Apps, the app runs you made, and the apps you published are handled on account deletion is now stated. 2026-10-12 service renamed (Assistant → everyagents). Amended on October 18, 2026: in Section 4, the retention of access logs changed from 30 days to 3 months (Article 41 of the Enforcement Decree of the Protection of Communications Secrets Act), and the retention of content-violation records changed from "as long as needed" to 3 years from the action. Section 5 now states how long the earlier steps of an edit conversation stay on Google's (7 days), OpenAI's (30 days), and xAI's (30 days) servers. Amended on October 11, 2026: the purposes and items of everyagents (Section 9-3) and per-reply Credit charge records were added, and the retention of everyagents conversations and AI usage records (deleted automatically after 3 months) was added to Section 4. Sections 5, 6, and 7 now state that conversation content is disclosed to, processed by, and transferred overseas to Google LLC to generate everyagents's replies. everyagents request records were added to access logs, and the right to delete individual conversations was added to Section 10. Amended on October 10, 2026: the AI Providers that receive generation requests (OpenAI, xAI, Black Forest Labs, BytePlus, and everyais platform community suppliers) and the USD merchant of record (Polar) were added as third-party recipients, and OpenAI, xAI, Black Forest Labs, BytePlus, and Polar were added as overseas recipients. International transfers are now listed by country, timing and method, purpose, and retention. The items and retention for API Keys and External Agents (Section 9-2) and the API Key safeguards in Section 11 were added. The items and retention of content-violation records, which are kept after account deletion, were added to Section 4. Amended on October 2, 2026: the items sent to Google, Meta, and X for advertising measurement now list product-usage events, member identifiers, hashed email addresses and mobile numbers, and IP and browser information at signup. Amended on September 26, 2026: X Corp. (X Pixel) was added as an advertising-measurement processor and overseas recipient.