Your work stays in your hands.

1. Website Library connection

The Library page requires Chrome, Context Container installed in the same profile, and an unlocked extension workspace. The site sends typed requests to the extension’s public ID. The extension accepts Library requests only from its configured exact website origin and only while the workspace is unlocked.

The site receives the same summary fields shown by the Library interface: Context name, source platform, creation and update dates, version number and count, turn count where available, and attachment metadata such as file name, type, size, availability, and whether a local copy exists. It can also request rename, remove-attachment, delete, upload, and shared-pin actions.

If you choose Upload for an attachment, the page handles the file you selected and transfers its bytes directly to the installed extension over a short-lived browser connection. The extension validates the transfer and stores the local copy in this Chrome profile. The file bytes are not sent to the website’s server or returned from the extension. Browsing the Library does not request attachment bytes. The website does not receive Context bodies, generated fields, tags, transcripts, or provider URLs, and it does not provide a server-side Context library, website profile account, cloud sync, or payment checkout. Hosted AI authentication is a separate service described below.

2. Information the extension handles

Capture starts when you choose it. The extension reads recognized visible user and assistant turns, visible pasted-text previews, attachment references, and source metadata such as platform, conversation title, URL, available conversation identifier, and capture time. It uses this information to generate a reusable Context and identify its source and version. New AI-enabled captures store the resulting structured Context and source metadata, rather than a new raw transcript. Older raw captures remain until you delete them.

The extension can temporarily retain files selected, pasted, or dropped into a supported chat composer so that an explicit Capture can preserve an original local copy when it can match that file to a visible attachment. This in-memory cache expires after five minutes and clears on page exit or a change to another existing chat. It is limited to ten files, 20 MiB per file, and 50 MiB in total. Reading pasted files does not read clipboard text or use the Clipboard API.

The extension does not intentionally collect passwords, hidden page state, unrelated browsing history, or content from unrelated websites. However, information you put into a visible conversation can itself contain personal information or secrets. Noise reduction is not a guarantee that sensitive information has been removed.

When you choose Insert or drop a Context onto a supported chat, the extension places that Context and any supported associated local files into the destination composer and activates its Send control after insertion completes. The destination AI service then processes what you send under its own policies. Opening the website, browsing the Library, and capturing a conversation do not send a chat message.

3. AI processing and other external services

The hosted AI build uses the Pordix API at api.pordix.com and the Google AI Studio Gemini API to generate Contexts. After the disclosure and your affirmative choice at sign-in, an explicit Capture sends selected, locally normalized conversation text, eligible visible preview text, extraction instructions and the requested output structure to Pordix. Pordix forwards that text to Google and returns the generated result to your extension. Original attachment bytes and provider attachment URLs are excluded. The API does not implement storage or logging of chat text, instructions, or generated responses; they are handled in operation memory. Private development builds can call the configured provider directly. This is separate from Google identity verification.

Google’s processing, retention, and data-use conditions depend on the service and billing configuration. Under its unpaid-service terms, submitted content and responses may be used to improve Google products and may be reviewed by people. Do not submit sensitive, confidential, or personal information to an unpaid service. Paid-service terms apply different data-use conditions; this policy does not promise a particular retention period or exemption from provider processing. See the Gemini API terms and Google Privacy Policy. A release using another extraction provider must disclose that provider before distribution.

Google sign-in is required for hosted AI captures; guest access remains available for the local Library. If selected, Google receives the requests needed for sign-in. The hosted extension sends the temporary Google ID token to Pordix over HTTPS; the server verifies its signature, audience, issuer and expiry. Pordix uses keyed pseudonymous identifiers derived from the Google account ID and verified email to apply sign-in access rules and usage limits, including any configured usage exemptions. A short-lived Pordix session token is returned to trusted extension session storage, which clears on browser restart; it is not stored in local Library records or exposed to content scripts. The verified email and sign-in time remain in extension local storage. Private direct-provider builds use a temporary token for the Google UserInfo check instead. Google tokens are not persisted by this application. Guest access stores a local workspace choice without an identity request or an email. Conversation and Context content are not included in sign-in requests.

Provider-hosted attachment downloads are off by default. Enabling the per-capture download option or choosing Save local copy for one attachment authorizes an eligible request to its source provider. The browser may attach its existing provider session credentials; the extension does not read or store cookie values. Downloaded bytes remain in the local attachment bundle and are not sent for AI extraction. Reference URLs may be stored locally for retry; the website Library does not receive them.

Optional sharing uses Firebase Authentication and Realtime Database for connection or delivery records. Nearby transfer exchanges connection metadata through Firebase and uses WebRTC to transfer the Context and files. STUN services operated by Cloudflare and Google help establish that connection and receive network addressing information. A peer may learn your network address as part of a direct connection.

Shareable links temporarily upload encrypted Context and attachment records to Firebase. The decryption key is in the link fragment. The handoff page can display a Context title supplied in that fragment and passes the link information to the installed extension; it does not display the Context body. Anyone who obtains the complete unexpired link may be able to receive its contents. The receiving browser keeps its own copy. See Firebase privacy information. Third-party services can process data in countries other than yours under their own terms and safeguards.

4. Storage, retention, and deletion

Contexts, source metadata, and attachment bundles are stored in browser-local IndexedDB. Settings, workspace access records, and pinned Context IDs use Chrome local storage. Saved records remain until you delete them or the browser removes the relevant extension or profile data. Sign-out removes the local access records; it does not delete your Context library. The website does not synchronize your Library between profiles or devices.

The normalized conversation stays in capture-operation memory and is discarded after that operation; it is not saved as a new raw transcript by AI-enabled capture. Temporary file-cache limits are described above. Sharing sessions have a ten-minute access window. The client attempts to remove relay records when a transfer completes or is cancelled. Expiry prevents access through the sharing protocol; it is not a guarantee that every provider record or backup has been physically erased at that moment.

Use Delete in the extension or connected website Library to remove a Context and its version history. Remove on an attachment detaches that attachment and its local copy from the selected version. Browser storage tools can clear local extension data. Removing the extension or profile can permanently remove local work, so retain an independent copy of information you need.

Local deletion does not delete a copy already sent to an AI provider, a chat service, or a sharing recipient. Requests concerning those copies must be made to the relevant recipient or service. Provider retention and any legal retention obligations are separate from local deletion.

5. Website requests and support information

Visiting a page sends your IP address, the requested path, request headers such as browser information, and request timing to the hosting infrastructure so it can deliver the website. Hosting infrastructure may record access and error logs for operation and security. These technical requests are separate from the extension’s local Library and are not a server-side copy of your Contexts.

The website application does not add advertising trackers or analytics scripts. Its hosted extraction API keeps hashed session tokens, pseudonymous account and email identifiers, expiry times, request and budget counters, and temporary concurrency records in a private server database for authentication and abuse prevention. It does not keep a cloud Context Library or a user profile directory. Sessions expire within one hour; minute counters and concurrency records expire after their short operating windows, and daily counters expire at the next UTC day boundary. Expired records are removed when the API next performs its normal cleanup; infrastructure backups, if configured, require separate retention controls. Signing out attempts to revoke the current server session and clears its extension token when local storage is available. If the server cannot be reached, that server session remains valid until expiry. The application does not use cookies to authenticate API requests or track visitors. Chrome and the third-party sites you choose have their own storage and cookie behavior. Connection requests to the extension do not upload your Library to the website host.

If you contact Pordix, your contact details and the information you choose to provide are used to respond, investigate the issue, and keep necessary correspondence. Hosting and email providers process the technical information needed to deliver those services. Send only what is needed for the enquiry. Do not send passwords, API keys, conversation contents, or attachment bytes in a support report.

Operational records and support correspondence are retained according to the time needed to resolve the enquiry, maintain service security, investigate incidents or disputes, and meet applicable legal obligations. These criteria apply to information Pordix holds; they do not set a retention period for your local Library or an independent AI provider.

6. Your choices and rights

You choose whether to capture a conversation, sign in, download attachments, upload a local copy, insert a Context, or share it. Guest access provides the local Library without Google sign-in; hosted AI captures require an enabled Google account, an unexpired service session and the processing disclosure at sign-in. You can stop future capture or sharing by not starting those actions, disable the extension, or uninstall it. Review the selected conversation before sending it for extraction. You do not have to provide content, but a feature cannot process material you do not supply.

You can browse local metadata, rename Contexts, remove attachments, and delete Context families. Generated Context fields cannot be edited in place; delete and recapture a Context if it needs replacement. Pordix cannot remotely inspect or erase a Library that exists only in your browser.

Under applicable UK data protection law, you may request access, correction, erasure, restriction, and portability of personal information Pordix holds, subject to the conditions and exceptions for each right. Where processing relies on consent, you may withdraw it by contacting us, and stop future optional processing through the controls described above. Withdrawal does not make earlier lawful processing unlawful.

Right to object: You may object to processing based on legitimate interests, including the service-security and support purposes described below. Contact support@pordix.com to exercise a right or raise a concern. Verification may be necessary to protect your information. A request to Pordix does not automatically remove data held independently by third-party providers or recipients.

You can also complain to the UK Information Commissioner’s Office through the ICO complaints service. Contacting Pordix first can help resolve the issue, but does not remove your right to complain to the regulator.

7. Limited use and security

Information received from Google identity APIs is used only to verify sign-in, show the active account, and authenticate and protect the requested hosted AI service, in accordance with the Google API Services User Data Policy, including its Limited Use requirements. Guest access does not use Google identity information. Context Container does not sell captured content or use it for advertising, unrelated profiling, or its own unrelated model training. This does not override the extraction provider’s terms described above.

The extension checks allowed website origins, validates transfer requests, limits attachment sizes, and encrypts shareable-link payloads before upload. Local browser storage is not a separately encrypted vault, and hiding generated fields in the interface is not encryption. People or software with access to your device, browser profile, or a complete sharing link may be able to access data. No storage or transmission method guarantees absolute security.

8. Eligibility and policy changes

The current AI-enabled service is intended for professional or business use by adults aged 18 or older. It is not directed to children. Do not provide a child’s personal information in a capture or support request.

This policy’s update date identifies the version you are reading. Changes to collection, providers, purposes, or sharing will be described here when introduced, with any additional notice or consent required by applicable law. Review this page when installing an updated release.

9. Purposes and legal bases

Where Pordix processes personal information to provide the product features you request, the legal basis is performance of the service agreement where that processing is necessary. Optional processing that requires consent will rely on that consent. For technical service operation, security, abuse prevention, and responding to general enquiries, Pordix relies on its legitimate interests in running a safe, reliable service and assisting users, balanced against your rights. Processing necessary to meet a legal obligation relies on that obligation.

Context generation is an automated summarisation feature. It is not used by Pordix to make decisions about you with legal or similarly significant effects. Do not submit special-category personal information, credentials, or material you are not authorised to process.

10. Publisher and privacy contact

Publisher: Pordix. Business country: United Kingdom.

For privacy enquiries, rights requests, security concerns, or product support, email support@pordix.com. Describe your request without including confidential conversation content, credentials, or attachment contents.

The current Privacy Policy is available at this Privacy Policy page; the Terms of Service describe use of the product.