This policy explains how Entevra handles personal information when you use entevra.com, the Entevra web chat, Community and the Entevra ADE (together, the services). Questions and requests go to support@entevra.com.

We use your information to run the services you ask for: signing you in, keeping your conversations and files, sending your messages to the AI model you choose, and keeping accounts safe and within their spending limits.

We do not sell personal information and we do not show advertising. Entevra does not train AI models on your content. The model providers that answer your requests process what you send under their own terms. For models reached through OpenRouter, we ask it to use only providers that do not collect it; models served by their own provider follow that provider’s terms (details below).

Account details: your email address, display name, account identifier, profile picture or chosen avatar, saved preferences and sign-in records. With Google sign-in we receive your Google account identifier, email address, name and picture; we ask only for the basic OpenID, email and profile scopes, which give no access to Gmail, Drive or Calendar. Email sign-in uses one-time codes, so Entevra never holds a password for you.

Your content: conversations and model responses, files you attach, images, video and audio you generate, memory notes you save, files, folders and memory notes you keep in a project, with the earlier versions of its text files, skills you write or import, conversations you share, and your Community posts, comments, reactions and mentions. Content can include personal information you choose to put in it.

Usage and spending: the models you use, token counts, request identifiers, costs, and your budget and limits.

Device and connection details: IP address, browser or app details and request times, recorded when your browser or the ADE connects to us. Our error diagnostics record fixed categories and status codes, not the text of your messages.

Entevra ADE sign-ins: when each computer signed in and when it was last active. We also keep the first and the last day each account used the ADE, to count active users; it is deleted with the account. The credentials the ADE uses stay on your computer; we store only one-way digests of them.

Support: the address you write from and whatever you include in your message. Please never send passwords, sign-in codes or provider keys.

To provide the services: sign you in, save and sync your history, send each request to the model you selected, generate the media you ask for, run web search when you turn it on or, with automatic web search on, when your message needs current information, and show your shared links and Community activity.

To keep accounts and the service safe: prevent abuse and fraud, investigate security problems, enforce the Terms of Service, and apply rate limits and monthly spending budgets.

To support you and keep you informed: send sign-in codes, answer support requests and tell you about important changes to the services. We do not send marketing email.

To keep the service reliable: we look at usage totals and error categories. We do not read your conversations to improve the product; the only people who can open account content are the administrators described below.

To meet legal obligations and to protect the rights and safety of our users, the public and Entevra.

OpenRouter routes your chat, image, video and voice requests to the provider that serves the model you choose, and the automated safety checks on images you post in Community to the providers of those classifiers. A request carries your message, the earlier conversation it needs for context, your display name (so the model can use it when you ask), any files you attach, and images from the conversation you choose to edit or to start a video from, the text of any skills you turn on for it and, when memory is on for that reply, the memory notes it uses. Model names, token usage and costs come back with the answer.

Our requests ask OpenRouter to use only providers that, under its policies, do not collect user data. That is not a promise of zero retention: providers may keep content for a limited time, for example to detect abuse, under their own terms, which can differ by model and change over time.

Models from other providers. A few models in the model picker are served by their own provider instead of through OpenRouter, and the picker shows that provider’s name. Today that is Abliteration (abliteration.ai). When you choose one of these models, our server sends that provider your message, the earlier conversation it needs for context, your custom instructions, and the chat’s project instructions and files if it has them. The provider receives no account details, and may keep request content under its own retention terms. We keep the reply in your conversation, like your other messages.

Your own provider keys. If you add your own key for an AI provider in Settings (an OpenAI-compatible service), we store it encrypted and use it only to check that it works, list its models, and send the requests you make with it to the provider you chose. Those requests carry your message, the earlier conversation it needs, your custom instructions, and the chat’s project instructions and files. They go to that provider under its terms, which may let it keep or use them, and your provider bills you for them. Replies on your key are text only for now: web search, Research, code execution and attached files are off for them, and they don’t read or change your memory notes. Titles, follow-up suggestions and predictions still run through our OpenRouter account, and summaries of long chats run on your key. We never show your key back in full; administrators see only that it exists, its provider’s host, its last four characters, when it was used and how much. Deleting the key or your account deletes it from our database; copies in backups stay encrypted and are kept as the retention section describes.

Fallbacks on your own keys. If you name a fallback for one of your models, and that model’s provider can’t answer before sending anything, Entevra sends the same request to the fallback model you named, through its provider and with your key for it. That provider then receives what the first one would have, under its own terms. The first provider may already have received the request. Nothing is sent to a model you haven’t named as a fallback.

You.com runs web search and You Search when you use them, and receives your search queries and the addresses of pages to read.

To show a source’s icon, our server asks that site for its icon. The site sees a request from Entevra, not from you, and receives no account details or message content.

ElevenLabs runs voice design when you describe a voice: it receives your voice description and the text to speak (or writes a short sample itself) and returns the audio (up to three voices from one description). We keep your description, the text and the audio in your conversation, like your other messages. ElevenLabs receives no account details, and may keep request content under its own retention terms.

Voice input. When you dictate in the web chat, your browser records your voice until you stop, and the recording is sent once through OpenRouter to a speech-capable model, which writes down what you said in the language you spoke. The text goes into your message box for you to review before sending, and we don’t keep the recording. Like your other requests, it goes only to providers that, under OpenRouter’s policies, do not collect user data.

File transcripts. When you attach an audio or video file in the web chat and choose Transcribe, your browser reads the file’s sound and sends it in parts through OpenRouter to a speech-capable model, which writes down what was said. The file itself, and a video’s pictures, stay on your device, and we don’t keep the sound. The transcript is attached to your conversation as a text file, like your other files. Like your other requests, it goes only to providers that, under OpenRouter’s policies, do not collect user data.

Custom instructions. If you write custom instructions in Settings → Personalization and keep “Use in new messages” on, we send them with each message you send in the web chat to the model you choose, in the same way as your message. They are not sent with Web search or Research replies, except automatic web search replies and a deep research’s report, as described below. We keep them on your account until you change or clear them, or your account is deleted; a long reply that is still being written keeps its own working copy until that reply’s data is cleaned up.

Projects. When you group chats in a project, we keep the project’s name, colour, icon and instructions on your account. If you use a project’s Files and folders, we also keep the folders and text files you create or import there, with each text file’s earlier versions. With each message you send in a project’s chat, we send the project’s name, its instructions and any instruction files you chose to use from its Files and folders, the names, kinds and sizes of its files and folders, the names and descriptions of the skills you chose to use, the titles of its other chats, the text of the files you added to it (or the parts most relevant to your message) and, when memory is on, its memory notes, to the model you choose, in the same way as your message and after your custom instructions; they are not sent with Web search or Research replies, except automatic web search replies, as described below; a deep research’s report is written with the project’s name and instructions only. When a reply needs it, the model can also search and read your other chats in that project; what it reads is sent to that model like the rest of the conversation. It never reads chats outside the project or in Trash, and you can turn this off in Settings. When a reply needs it, the model can also search and read the text files in the project’s Files and folders and the skills you chose to use; what it reads is sent to that model like the rest of the conversation and counts toward that reply’s cost. This is on by default, and you can turn it off in Settings (“Let replies read your project files”). Files and folders you mark “Never send to models”, or that your project’s .entevraignore file names, are never sent. A reply can also propose changes to the project’s text files (edits, new files, renames or moving files to Trash); nothing changes until you apply a change on the card under the reply, and each applied change keeps the file’s earlier version. Deleting a project deletes its name and instructions; its chats stay in your history unless you choose to move them to Trash. Imports and exports: when you import into a project, from a ZIP file or a public GitHub repository (our server downloads it from GitHub, which receives that request, not your account details), the files are kept on our servers for up to 24 hours while you review them and are deleted if you don’t complete the import. When you export a project, we build the file as you download it and don’t keep a copy. A long reply that is still being written keeps its own working copy until that reply’s data is cleaned up. Projects are deleted with your account.

Helper requests. To give a new chat its title and to suggest follow-up questions after a reply, we send the start of your message and of the reply through OpenRouter to a small model we choose. When you set a goal for a chat, the model you choose writes its plan, and after each step a small model we choose reads the goal, the plan and that step’s reply to check progress. Like your other requests, these go only to providers that, under OpenRouter’s policies, do not collect user data. We keep the title, the suggestions and the goal’s progress with the chat, and they are deleted with it.

Predictions as you type. When predictions are on (Settings → AI behavior), the web chat sends the message you are still typing, with your previous message and the start of the reply to it, through OpenRouter to a small model that suggests how to finish your sentence. Nothing is sent until you pause typing, and only providers that, under OpenRouter’s policies, do not collect user data are used. We don’t keep your unsent text or the suggestion; we keep only a count of how many predictions you used each day. You can turn predictions off in Settings at any time.

Predictions. When predictions are on (Settings → AI behavior), after a reply finishes, or when you click into the message box of a chat that ends with a reply, the web chat sends the latest part of the conversation (your last two messages, the start and end of the reply, and any follow-up suggestions on it) through OpenRouter to a small model that suggests the message you might send next. Only providers that, under OpenRouter’s policies, do not collect user data are used. We hold the suggestion in server memory for up to 10 minutes, so reloading the page doesn’t ask again, and never save it or the text it was made from; we keep only a count of how many predictions you used each day. You can turn predictions off in Settings at any time.

Automatic web search. Automatic web search is on unless you turn it off in Settings → AI behavior, or for one chat in its + menu. Before a reply in the web chat, our server may send your latest message and your previous one through OpenRouter to a small model that decides whether the reply needs current information from the web. When it does, the search uses only messages you wrote: the model you chose (or, for a model that can’t use tools, a small model we choose) writes the search queries from them and from what its earlier searches return, You.com runs the searches, and pages found, or addresses in your message, may be opened and read. Your memory, custom instructions, project context, files and the replies in the chat are never used to search. The reply then answers with the sources it found, and with your instructions, project context and memory as a reply without search would. These requests go only to providers that, under OpenRouter’s policies, do not collect user data. The searches, the pages read and the model’s work to run them are charged to you, like a Web search reply; the check and the query writing for models without tools are free. We keep the searches a reply made, which you see in its details.

Looking up past chats. If you turn on “Let Entevra look up your past chats” in Settings, the model answering a chat outside a project can search your other chats outside projects and read parts of them when your question needs it; the reply shows when it did. It never looks at temporary chats, chats in Trash or other members’ chats, and it is told to look at archived chats only when you ask about them. What it reads is sent to the model you chose for that reply, like the rest of the conversation; the reply keeps a note of what it searched for and which chats it read, which you see in its details. A long reply that is still being written also keeps what it read in its own working copy until the reply ends, or for up to 7 days while it waits for you to continue it; nothing else is stored. It is off until you turn it on, and you can turn it off at any time.

Deep research. When you start a deep research, our servers run it until its report is written, even if you close the page (it isn’t offered in a temporary chat). From your message, up to two of your earlier messages in that chat, the plan you approve and any changes you ask for before it starts, and then from the pages it reads (never from your memory notes, files, other chats, custom instructions or project instructions), it writes web search queries for You.com and chooses pages from the results for You.com to read. It sends those messages, the plan, the search results, the text it read and its notes through OpenRouter to the model you chose, and we ask OpenRouter to use only providers that do not collect it; your custom instructions and your project’s name and instructions are sent only to write the report. We keep the queries, the results, the text it read, its notes and its progress while it runs and for up to 7 days after it ends or pauses; the report (or, if it ends early, its plan or its notes so far), its sources, and the searches and page reads it made stay in your conversation. Deleting the conversation permanently deletes them at once, and so does deleting your account. Writing the plan, and each new plan you ask for, is charged like a short reply; each search and page read is then charged at Entevra’s price for it, with the model’s use, within the budget shown before you start, and if the report needs more, it stops and asks before going past it. Failed searches and page reads are not charged, and a research that finds nothing it can read stops without writing a report.

Code execution. When you turn on Code execution in a chat, the model answering can run commands in a sandbox that OpenRouter runs for that reply, with no network access. The sandbox receives only the commands the model writes, which can include text from your conversation; the commands and what they print are shown in the reply and kept with the conversation, like the rest of the reply. Files the commands create stay in the sandbox, where OpenRouter keeps them for up to 30 days after it was last used. We fetch and keep only the kinds of file described under “Files a code run makes” below; we don’t fetch or keep any others. The sandbox’s running time is billed with the reply. Like your other requests, the model’s part of a run goes only to providers that, under OpenRouter’s policies, do not collect user data.

Files a code run makes. When a code run writes a document, spreadsheet, presentation, PDF, image, CSV, text, Markdown or JSON file in OpenRouter’s sandbox and the reply points to it, we fetch it once and keep a copy in your private storage, with its name and its path in the sandbox. We check each file first, and we don’t keep files with macros. Scripts, web pages and other kinds of file stay in the sandbox and are only named in the reply. Kept files count against your storage allowance and appear in the chat and in Settings → Files; keeping them doesn’t change what the reply costs you. Files from temporary chats aren’t kept. Deleting the chat (when it leaves Trash) or your account deletes them, and their stored copies are removed from our private storage in a later cleanup.

Notifications. We keep a list of notifications on your account: when someone comments on your Community post, replies to your comment or mentions you; when a reply you started finishes, stops early or asks you a question after you’ve left the page; and when a goal you set is met, can’t be met, or pauses because its budget or your allowance ran out; and when an administrator replies to a message you sent them. We keep at most 200, for up to 30 days. Each one says what happened and links to it, with the display name of the member who acted; it doesn’t copy the content. When you comment, reply or mention someone in Community, they get a notification with your display name.

Scheduled tasks. When you schedule a task, we keep its prompt, schedule and settings. At each scheduled time, while you’re away, we send the prompt with the task’s last prompts and answers, your custom instructions and, if you turn it on for the task, your saved memory, to the model you chose, and to the search provider when Search is on, and keep the reply in the task’s conversation. A run never reads your other conversations. Each run is charged to your monthly allowance like your other replies, up to the task’s cap; a run Entevra couldn’t start may be tried once more ten minutes later. A task pauses by itself when its cap is too low, its model or web chat is no longer available to you, its conversation is in Trash, three runs in a row fail, or 30 days pass without you opening Entevra; it ends when its conversation is deleted for good. Deleting a task stops it and removes its prompt at once; its conversation stays until you delete it.

Connectors. If you connect a service in Integrations → Connectors, you sign in to that service on its own page, and it gives Entevra a token that lets the model answering your chats use the tools you leave on. We keep the token encrypted, use it only for your chats, and never show it to anyone, including administrators. When a reply uses a connector, our servers send that service the request the model makes, which can include text from your conversation, and what the service returns goes to the model you chose for that reply, like the rest of the conversation. We don’t keep what a service returns on its own, except in a long reply’s working copy, which keeps what it asked a service and what the service returned only as long as described under Long replies, but the reply and the model’s reasoning can quote or summarize it, and they are kept with your conversation like any other reply, including in links you share. The reply also records which tools it used and a short note of what it asked. Each service handles your data under its own terms. Disconnecting, or deleting your account, deletes the token, and we ask the service to revoke it where the service supports that. Connectors are off until you add one. If you ask to be told when a service can be connected, we keep that request until we tell you, you cancel it or you delete your account; administrators see only counts for each service (how many asked, how many were told and the day of the latest request), never who asked.

Supabase stores our database and private files; our primary region is in the European Union. Hostinger hosts and delivers the website and application. Google provides Google sign-in, and Google Workspace handles email sent to support@entevra.com. Resend delivers email sign-in codes and keeps delivery records such as the recipient address. GitHub supplies the files when you import a skill from a GitHub address, and receives that address from our servers.

Sign-in protection. When you sign in or create an account, Cloudflare Turnstile checks that the request comes from a person rather than an automated program. It runs small checks in your browser and receives technical information such as your IP address and browser details. It does not receive what you type into the sign-in form, and we receive only the result of the check. Cloudflare processes this information under its Turnstile Privacy Addendum, linked below.

These providers and their own suppliers may process information in countries other than yours. A chosen region does not mean that every copy or operation stays there. If your content needs specific confidentiality or residency, review the provider pages below before sending it.

OpenRouter provider data policies

OpenRouter privacy policy

ElevenLabs privacy policy

Supabase privacy policy

Cloudflare Turnstile Privacy Addendum

Hostinger privacy policy

Google privacy policy

Resend regions and data residency

GitHub privacy statement

A skill is a set of written instructions that you turn on for a message. Skills you write, paste, upload or import are stored with your account, together with each version you save, and only you can use them.

When you import a skill from a GitHub address, our servers read the files from that public repository for you and store them as your skill. GitHub receives the request from Entevra, not from your browser.

Each skill is checked automatically when you save or import it, for instructions that try to override Entevra’s rules or hide things from you, requests for passwords or payment details, commands that download and run software, and hidden or encoded text. You see the results. These checks run on our servers; they do not send your skill to anyone.

When a skill is on for a message, its text goes to the AI model with that message, in the same way as the message itself. We record which skills each reply used, and we count how often each library skill is used.

Skills in the library are written by the authors named on them and published under their own licences. “Reviewed by Entevra” means we read and tested the skill before adding it, not that we guarantee its results. Legal and finance skills give general information, not professional advice.

Skill suggestions while you type are worked out in your browser; your draft is not sent to us for them. You can turn them off in Settings.

If you report a skill, we keep the report and the reason you give, and administrators can see which account sent it. Deleting one of your skills moves it to the Deleted list on your Skills page, where you can restore it. To delete a skill permanently, contact support@entevra.com.

A shared conversation is a copy that anyone with its link can open. A new link shows your display name and picture unless you turn that off when you share it. Settings → Sharing sets the starting choice for new links, and you can change it on each link later. Your email address is never used as the name on a link.

You can revoke a link at any time from the conversation or from Settings → Sharing. A link stops opening while its conversation is in Trash, and deleting the conversation permanently deletes its links.

A signed-in member who opens a live link can continue the chat in their own account. They get a copy of what the link shows: the title, the messages, and each reply’s model name, reasoning and time. Your name, picture, files and account details are never copied. The copy is theirs: it stays in their account after you revoke the link, delete the conversation or delete your account.

Community is visible to signed-in members. Your posts, comments and reactions appear with your display name and avatar. Moderators can hide content that breaks the rules.

Images you post in Community are visible to members who can see the post or comment. Before an image appears, it is checked for safety by automated classifiers run through OpenRouter; images that fail the check are not posted. Images are removed when you delete the post or comment, and administrators can remove images that break the rules.

Messages to administrators. When you message Entevra’s administrators from Community, we keep your messages and their replies with your account so that they can answer you. Only administrators can read them, and each time one opens a message it is recorded in an audit log with the administrator, the account and the time, as for your other content. Replies are signed “Entevra”, not with an administrator’s name. You can close a conversation with the administrators but not delete it: it is kept while your account exists and deleted with your account. Your messages aren’t shared with anyone else or used to train models.

Apart from the providers above and the sharing you choose, we disclose personal information only when the law requires it, to protect people from harm or to protect the services, or as part of a merger or sale of the business, in which case this policy would continue to apply to your information.

A small number of authorized Entevra administrators can open an account’s details, usage, conversations, files, skills and shared links when needed to provide support, investigate abuse or security problems, or meet legal obligations.

Every time an administrator opens account content, the view is first recorded in an audit log with the administrator, the account, what was opened and the time. Administrators cannot change or delete those records.

Administrators can also suspend an account or change its email address. A suspended account can still sign in and read its history; only new chat is locked. Changing an account’s email signs out its existing sessions.

When needed for support, abuse or security problems, or a legal obligation, administrators can also move an account’s conversations to Trash, restore them, delete them permanently, remove files and revoke shared links. Each of these changes is recorded in the audit log before it is made, with the administrator, the account, what was changed and the time. A permanent deletion is recorded by the conversation’s identifier, not its content.

You sign the ADE in by approving it in your browser on entevra.com. The ADE never sees your password or sign-in codes. Its requests go to agent.entevra.com and use the same account, history, limits and providers as the web chat.

An ADE request carries the messages, code, file contents and tool results that you or the coding agent include in it. Your projects otherwise stay on your computer. Agent runs never send your uploaded private files or memory notes.

If you connect your own provider key in the ADE, those requests go from your computer directly to that provider under your agreement with it, and Entevra does not receive them.

Settings → Account → Entevra ADE lists the computers signed in to your account and signs the ADE out everywhere.

If you used our earlier chat service, we moved your account identifiers, verified sign-in links, preferences, conversation history, usage records and files so you can continue with the same account. Sign-in links are checked before access is granted; accounts are never merged because names match.

We keep a restricted copy of the migrated records to check completeness and recover from errors. A file that a security check flagged stays listed but cannot be opened. Contact support if history, an account link or a file seems wrong.

We keep account details and content while your account is active, so your history stays available. Deleting a conversation moves it to Trash, where you can restore it for 30 days; after that it is deleted permanently, and Trash shows how many days each conversation has left. You can also delete a conversation in Trash permanently sooner, one at a time or with Empty Trash. Trash is not emptied automatically while an account is suspended.

Temporary chats. A temporary chat isn’t listed in your history or in search, can’t be shared, exported or moved into a project, and doesn’t read or change your memory notes. When you leave it, we delete it permanently with its messages, attached files and generated media; if that can’t happen right away (for example, you went offline), it is deleted 24 hours after your last message in it. While an account is suspended, its temporary chats are kept until the suspension ends. Its messages are processed by the model you choose, as in any chat, and usage and spending records for its replies are kept as described below, without its content.

Deleting a conversation permanently deletes its messages, the files attached to it, the images, audio and video generated in it, and its shared links, and it can’t be undone. Attached files and generated images can’t be opened from that moment, and their stored copies are removed from our private storage in a later cleanup. Memory notes are kept separately and are not deleted with a conversation. Files, folders, earlier versions and memory notes you keep in a project are stored privately until you remove them (an item you move to the project’s Trash stays there for 30 days) or delete the project, and are deleted with your account.

Long replies. A long reply may keep running on our servers after you close the page or lose your connection (except in a temporary chat), and it ends in your conversation. To resume it after an interruption, we keep a working copy of what it was given and of its progress while it runs and for up to 7 days after it ends or pauses, and its streamed text for a day after it ends (or as long as it waits for you to continue it, at most 7 days); deleting the conversation permanently deletes them at once, and so does the deletion of your account. A long reply is charged to your monthly allowance like any reply and stops for your go-ahead before it passes $4; each time you choose Keep going, it may spend up to another $4.

Deleting your account. To delete your account, contact support@entevra.com or send a request from Settings → Account. An administrator schedules the deletion and writes a note you’ll see when you sign in. Until the date shown, your account is closed, but you can still sign in to read your conversations and export them, and an administrator can restore the account if you ask. On that date we permanently delete your conversations and their files, your projects with their files, folders and earlier versions, memory notes, the skills you wrote, shared links, your Community posts, comments and likes, your messages to administrators, your notifications, your preferences and your profile, and your email address is freed. Stored copies of files are removed from our private storage in a later cleanup. Usage and spending records, security logs and administrator audit records are kept as described below, without your content.

Browser sessions stay signed in for 30 days after you last use them, and for at most 90 days after you sign in. Email codes expire after eight minutes and Google sign-in attempts after five. ADE access lasts 15 minutes at a time and renews for up to 30 days, unless you sign the ADE out sooner.

Usage and spending records (such as the model, token counts and cost of each reply), security logs, administrator audit records, backups and migration copies are kept longer for accounting, security and recovery, and deleting a conversation does not remove them. We have not yet set a fixed automatic deletion schedule for all of them, so we review each deletion request against the records involved and any legal obligations.

Signing out or clearing your browser does not delete records on our servers, and copies already processed by a provider follow that provider’s retention rules.

Sign-in uses essential, secure, HttpOnly cookies for your session and for short-lived Google or email sign-in attempts. Signing out ends the current session. Without these cookies you cannot sign in.

After you sign in with an emailed code, we keep one more essential, secure, HttpOnly cookie on that browser for 180 days. It holds a keyed fingerprint of your email address, not the address itself, and lets that browser keep receiving sign-in codes even when someone else has asked for many codes for your address.

Your browser’s local storage keeps preferences such as your theme, and the web chat keeps an unsent draft per account in the tab’s session storage so it survives a reload. Clearing site data removes them.

We use no advertising or analytics cookies. Services you visit directly, such as Google during sign-in, may set their own cookies under their policies.

You decide what to send to a model, whether memory and web search are used, what to share and what to post in Community. You can change your name and picture in Settings, choose whether shared links show them, export a conversation from its menu, revoke shared links, and delete conversations permanently from Trash.

Depending on where you live, you may have the right to access, correct, delete, export or restrict the use of your personal information, to object to its use, and to withdraw consent. To make a request, write to support@entevra.com from the email address on your account. We may need to confirm that the account is yours first.

If you think we have not handled your information properly, tell us. You may also be able to complain to your local data protection authority.

Email Entevra privacy support

Connections are encrypted. Sessions and sign-in attempts expire, files sit in private storage and are checked against your account before every delivery, provider credentials never leave our servers, and the ADE’s credentials are stored by us only as digests. Administrator access to content is audited, and spending limits cap what any request can cost.

No system is perfectly secure. If a security incident affects your information, we will investigate it and notify you and the authorities where the law requires.

The services are for people aged 16 and over. If you believe a child under 16 has given us personal information, contact support@entevra.com and we will review and remove it.

We update this policy when the services or the way they handle information change, and show the date of the latest version above. We will tell you about significant changes in the services before they take effect.

Entevra provides entevra.com, the Entevra web chat with access to third-party AI models, image, video and voice generation, sharing, Community, and the Entevra ADE, a desktop app for coding with AI (together, the services). These terms are an agreement between you and Entevra. By using the services, you accept them and the Privacy Policy.

You must be at least 16 to use the services. If you use them for an organization, you confirm that you may accept these terms on its behalf.

Give accurate account information and keep your email and Google account secure, because they are how you sign in. You are responsible for activity on your account; tell us at support@entevra.com if you think someone else has used it.

Do not share an account or create extra accounts to get around limits. We may suspend or close accounts that break these terms or put the services or other people at risk. A suspended account keeps read access to its history unless the law or safety requires otherwise.

Each account has a monthly usage budget. A request reserves part of it while it runs and is then charged at its actual cost; when the budget is used up, new requests pause until it renews. Free models depend on their providers and can have their own limits.

Plans and prices are shown on the Pricing page. There is no paid checkout today. If paid plans open, the price, billing, renewal, cancellation and refund terms will be shown before you pay.

We may set rate limits and other limits to keep the services fair and stable.

Do not use the services to break the law or to help anyone else break it.

Do not create or share content that sexualizes minors, promotes violence or terrorism, or harasses, threatens or exposes other people.

Do not build or spread malware, attack systems, or probe, overload or get around the security, limits or budgets of the services.

Do not impersonate others, mislead people about AI-generated content in harmful ways, or infringe anyone’s intellectual property or privacy.

Do not resell access to the services, scrape them, or use them to build a competing service.

Follow the usage policies of the model providers whose models you use; they apply to your requests too.

You keep your rights in what you submit. You give Entevra permission to host, copy, process and transmit it only as needed to run the services for you: to store your history, send requests to the providers you choose, and show your shared links and Community posts to the people you share them with.

You confirm that you have the rights and permissions needed for what you submit, including any personal information about others. We may remove content or limit features when content breaks these terms or the law.

AI models can be wrong, out of date, biased or entirely made up. Check anything important before you rely on it. Output is not legal, medical, financial or other professional advice.

As between you and Entevra, you may use the output you receive, subject to the law and the model provider’s terms. Output is not unique: other people may receive similar results, and some output may not be protected by copyright.

You are responsible for how you use output and for the decisions you make with it.

The ADE’s coding agent can read and change files and run commands on your computer within the permissions you choose. Review its changes and commands, choose permissions carefully and keep backups; you are responsible for what it does on your systems.

The ADE is in preview. Features can change, and some actions may not ask for confirmation, depending on the mode and permissions you choose.

Anyone with a shared link can view the shared copy until you revoke it. A signed-in member who opens it can also continue it in their own account with a copy of what it shows, which they keep after the link is revoked. Share only what you are comfortable making visible.

In Community, be respectful, stay on topic and do not post spam or other people’s private information. Moderators may hide posts or comments and limit accounts that break these rules.

The AI models available through Entevra are built and run by third parties, reached through OpenRouter or, for a few models, directly from their own provider. Which models are offered, how they behave and what they cost can change when a provider changes or retires them.

Entevra does not control those providers and is not responsible for their services, though we choose how requests are routed as described in the Privacy Policy.

We keep improving the services, so features may be added, changed or removed. Some features are previews and may be less reliable than others.

We work to keep the services available but do not promise that they will be uninterrupted, error free or available in every location.

To the extent the law allows, the services are provided as is and as available, without warranties of any kind, including warranties of merchantability, fitness for a particular purpose and non-infringement.

To the extent the law allows, Entevra is not liable for indirect, incidental, special, consequential or punitive damages, or for lost profits, revenue, data or goodwill. Entevra’s total liability for any claim relating to the services is limited to the greater of the amount you paid Entevra for the services in the 12 months before the claim and US$100.

Nothing in these terms limits rights or remedies that cannot be limited under the law that applies to you, including consumer protection rights.

You can stop using the services at any time and ask support@entevra.com to delete your account.

We may suspend or end your access if you break these terms, if your use creates risk or legal exposure for others or for Entevra, or if the law requires it. Where it is reasonable, we will tell you why. Sections that by their nature should continue, such as content permissions, disclaimers and liability, continue after your use ends.

If you have a concern, contact support@entevra.com first so we can try to resolve it informally.

These terms are governed by the law of the place where Entevra is established, without regard to its conflict-of-law rules. If you are a consumer, you also keep the protection of the mandatory laws and courts of the country where you live.

We may update these terms as the services change. The effective date above shows the latest version, and we will give notice in the services before significant changes take effect. If you keep using the services after that, the updated terms apply; if you do not agree, stop using the services.

If any part of these terms cannot be enforced, the rest still applies. Questions about these terms go to support@entevra.com.