Doodle Charts: stateless rendering, device-only history
Trustflow operates Doodle Charts without accounts, analytics, advertising identifiers, a user database, or server-side history. Attach original files directly in ChatGPT; the current ChatGPT model reads and analyzes them under OpenAI’s policies. Original files are not uploaded to Doodle Charts.
The remote MCP receives structured chart values, labels, source filenames and AI analysis solely to validate the requested chart or dashboard and return the result in the same response. These request data exist only during request processing in memory. We do not retain requests, results, user profiles, files or history, write them to disk, back them up, or share them with another model API. Application request logging and access logging for the Doodle Charts endpoint are disabled. Network connections necessarily process IP addresses in transit; they are not retained by this application.
The app saves chart and analysis history in IndexedDB on the current device, and language preferences in local browser storage. This storage belongs to the client’s app origin, is not account authentication, is not synced to our server, and may be unavailable or cleared by the client. Other people using the same device and app storage may be able to see that history. Use Delete or Clear history, or clear this app’s client storage, to remove it.
HTML, PNG, SVG and JSON exports are generated on your device. Supported clients receive the file directly through their download interface. The compatibility page carries a compressed export only in the URL fragment, which is not included in HTTP requests; it removes the fragment from browser history after loading. There is no server export relay. Keep or delete downloaded files separately, and avoid sharing the private download URL. ChatGPT retains its own chats, attachments and tool responses under its policies; deleting device history does not delete those copies.
Local deletion also records the removed random chart identifiers, without filenames, values or analysis text, to prevent old ChatGPT tool responses from automatically restoring deleted records. These device-only markers are removed when you clear this app’s client storage.
Fusion Reader: anonymous RSS reading with a local JSON library
Fusion Reader is operated by Trustflow and is covered by this policy. It does not require an iPlugin account, Google sign-in, OAuth authorization or installation of a local service. The account, PostgreSQL canvas and Google sign-in descriptions below apply to account-backed products such as Offshoot, not to Fusion Reader.
Data categories, purposes and retention
| Information | Processing and purpose | Retention |
|---|---|---|
| Subscriptions, article history, bookmarks, settings, translations and article conversations | The Reader stores the personal library as JSON in the device-local storage space supplied by the ChatGPT client. OPFS is preferred, with a limited localStorage fallback. The remote MCP does not maintain a personal library database. | On the device until removed through available Reader or host storage controls. Clearing the host plugin storage can erase it. Exported JSON files remain until you delete them separately. |
| Public RSS and article URLs, source titles and fetched public content | The remote MCP at https://iplugin.dev/fusion-reader/mcp receives a URL, reads the public source and returns parsed content. The source publisher sees a network request from our service. Private, credential-bearing and internal-network feeds are unsupported. | The application processes these requests in memory for the current operation; it does not save a per-user subscription list or article archive on the server. |
| Translation text, article questions and answers, settings changes and action identifiers | Requested save operations pass their payload through the remote MCP for validation and return it to a client delivery card. Fields can include client article/feed IDs, paragraph citations, source hashes, request/delivery IDs, feed labels, language and reading preferences. The existing Reader applies the action to its local JSON library. | No persistent server record of these action payloads is created by the application. Delivery actions can remain in the device-local inbox until the Reader applies them or the host storage is cleared. |
| Context shared with ChatGPT | The Reader supplies subscription names/URLs and state, selected article identifiers and source paragraphs, language preferences, article counts and applied request IDs so ChatGPT can perform the requested workflow. Explicit AI requests include the selected source and relevant article conversation context. | OpenAI handles this context and linked conversations under your ChatGPT settings and its own policies. Deleting the Reader library does not delete the ChatGPT conversation. |
| Connection and security metadata | Our hosting infrastructure handles IP addresses, connection times, HTTP paths/statuses and errors. The application uses in-memory rate-limit counters and does not intentionally log MCP request bodies. Ordinary operational logs are separate from the personal library. | Rate-limit state is transient process memory. Infrastructure logs follow the hosting system's rotation and security-retention configuration; no fixed number of days has been independently verified for this notice. Contact Trustflow for the current applicable period or a request about identifiable operational records. |
Who receives information
Trustflow operates the remote service. Its hosting providers process connections and transient requests needed to run it; authorized operators may access operational records for support and security. OpenAI processes ChatGPT context and model requests separately. The public RSS/article publishers you choose receive requests from the remote fetch service. Following a source link or loading external source media can also contact that publisher or media provider directly. We do not sell the Reader library, provide it to advertisers or use it for an advertising profile. Remote hosting, OpenAI and source publishers may process information outside your country under their respective arrangements.
Your controls and deletion options
You can choose which public feeds to add, pause and restore subscriptions, change reading preferences and export a local JSON backup in Reader Settings. Backups are not uploaded to the MCP. Automatic translation is off by default. The current host does not provide silent model execution; manual AI requests appear in the linked ChatGPT conversation. Automatic refresh is initiated by the client only while the Reader is open and visible, with overdue checks resumed on reopening.
There is no cross-device library synchronization or Fusion account. Different host storage spaces can contain separate libraries. Clearing the Reader's plugin storage through your ChatGPT client removes that space's library and pending actions; uninstalling the plugin alone may leave storage behind. Delete exported JSON copies separately and manage ChatGPT conversations and OpenAI data controls separately. Trustflow cannot retrieve or remotely erase JSON that exists only on your device. See the Fusion Reader deletion instructions.
For questions, access/correction/deletion requests about information Trustflow actually holds, or to raise a privacy concern, contact micromvp@gmail.com. Where applicable, you may also request restriction or object to processing and complain to your data-protection authority. Provide only the minimum information needed to locate an operational event; do not send your full library, private article content, passwords or ChatGPT credentials. Reasonable verification may be needed to protect another person's records. We update this policy's date when the described processing changes.
01. Scope and approach
iPlugin is a collection of plugins for ChatGPT. Our first plugin, Offshoot, organizes content into an editable, branching canvas. We process information needed to provide sign-in, authorization, storage, security, and service maintenance.
Content is generated by the model in your ChatGPT conversation. iPlugin does not need your ChatGPT password, ChatGPT token, or a model API key. OpenAI’s policies govern how ChatGPT handles your conversations.
02. Information we process for account-backed services
The following account and canvas categories apply to Offshoot and shared sign-in. Fusion Reader has no account and uses the separate data flow described above.
| Information | Purpose |
|---|---|
| Google account ID, email, and name | Identify your account, establish an iPlugin session, and link project access. |
| Operational usage records | Record first Google sign-in time and successful node-generation events (time, account ID when available, and identifiers used to avoid counting retries twice). Restricted administrators use these records to view daily registration, active-user and generated-node totals. Event records do not contain node content and remain when canvas content is deleted; account-data requests can include these records. |
| Display name and uploaded profile photo (optional) | Show your chosen iPlugin profile. Photos are resized in your browser and stored with your account in PostgreSQL; editing this profile does not change your Google account. You can replace or remove your photo in plugin settings. |
| Canvas and artifact content | Store prompts submitted to the plugin, nodes, writing, webpage source, artifacts, and editing history. |
| Authorization records | Register MCP clients, manage consent, refresh tokens, bind workspace access, and perform security checks. |
| Technical information | Handle connections, rate limiting, and troubleshooting. Web server logs may include IP addresses, request times, and paths. |
| Browser cache and preferences | Recover working drafts and remember language, theme, and other interface preferences. |
03. Google sign-in
We request only openid, email, and profile to verify your identity. This does not give us access to Gmail, Google Drive, contacts, or calendars.
After verifying your Google identity, the server creates an iPlugin session. We do not store Google access or refresh tokens. You can manage or revoke Google access in your Google account’s third-party connections.
Google sign-in verifies who you are. MCP authorization determines which client you allow to access your personal Offshoot workspace. Each Google account has separate canvases, generation history, trash settings, and file access.
04. Storage, sharing, and third parties
Cloud canvases, artifacts, revision history, generation queues, and authorization records are stored in PostgreSQL on our VPS. Your browser may also retain a working cache. The optional local stdio version stores data on the device running it.
Each Google account has its own isolated Offshoot workspace. Canvases, history, trash settings, and file access are scoped to your verified account and are never shared between accounts. Authorizing a client grants access to your workspace only.
Content submitted through MCP travels between your client and iPlugin. Google provides Google sign-in, and ChatGPT provides its conversation and model services. Each provider processes information under its own policies.
We do not sell personal information or provide canvas content to advertisers. This website currently uses no advertising trackers or marketing analytics cookies. Loading external media in an artifact may send a network request to that media provider.
05. Retention and backups
Canvases are kept until you delete them. Trash does not automatically empty by default. You can choose a retention period of 1, 7, 30, 90, or 180 days. Permanent deletion removes associated content, snapshots, and generation queues. A minimal deletion marker remains to prevent an old cache from restoring deleted content.
We retain only the five most recent successful, validated database backups. Deleted information may remain in an older backup until it is replaced by a newer successful backup. Failed backups do not trigger deletion of existing backups.
Login requests generally expire after 10 minutes, project handoff tickets after one minute, and iPlugin browser sessions after seven days. MCP access tokens generally last one hour; refresh tokens last up to 30 days and rotate when used. Your personal workspace is associated with your verified Google account identifier.
06. Your choices
You can export, edit, move canvases to trash, permanently delete them, or disconnect your MCP client. Browser caches must be cleared on each device. Disconnecting or signing out does not automatically delete canvases.
To request account deletion, remove a Google binding, or revoke all issued tokens, follow our data deletion instructions. Requests affecting a personal workspace require verification of account ownership.
07. Contact and updates
iPlugin and its plugins are operated by Trustflow. For privacy questions or requests, contact micromvp@gmail.com or use the support page. In public feedback, describe the type of request without sharing keys, tokens, or private canvas content.
We will update this page and its revision date when our service or processing practices change. Material changes will be communicated through available product channels.
Files and Cloudflare R2
When enabled by the operator, uploaded files, profile images, and generated media are stored in a private Cloudflare R2 bucket. File metadata and durable references are stored in PostgreSQL. Preview and download links expire after one hour; anyone holding a link can access that file until it expires. Fresh download links are issued only to the account that owns the file.
Database backups contain file references and metadata, not R2 file contents. Files may remain in R2 after a node or canvas is removed; request permanent file deletion from the operator. Downloaded copies and previously issued links may remain available independently.
QR Studio
QR Studio processes the text, links, Wi-Fi details, contact information, images and other content you choose to encode, along with your design settings. The hosted service receives this content to render, save and export your QR code. OpenAI processes information in your ChatGPT conversation under its own policies, and our hosting provider processes network requests and operational logs.
Hosted designs are kept in the current server-issued MCP session. Session libraries and server exports are removed after seven days, and may become unavailable sooner if a session ends or the service restarts. Anyone with an export URL can access that file until it is removed. Language and theme preferences may be stored in browser storage. Local installations store their own designs and exports on your device.
You can delete a design in the studio and follow our data deletion instructions. Do not share QR codes or export links containing passwords or other information with people who should not have access to them. For privacy questions contact micromvp@gmail.com.
Resume Optimizer: client-local resumes
Resume Optimizer is operated by Trustflow and requires no iPlugin account. You may provide contact details, employment and education history, skills, photographs, job requirements, and writing preferences. The editor uses them to prepare and customize your resume. PDF and DOCX text extraction and image OCR run on your device; OCR code and only the needed language data are downloaded as static assets. Selecting a resume in the normal import flow does not upload its original file to our server.
When you ask ChatGPT to generate or improve a resume, the extracted text and your instructions are shared with the current ChatGPT conversation. Tool calls carrying resume sections and explanations pass transiently through our hosted MCP relay. The relay has no resume database and does not retain resume bodies or write them to application request logs. Hosting and network providers process technical connection data, such as IP addresses and request timing, under their own operational practices; we have not independently verified every provider retention period. OpenAI processes conversation content under its own policies. We do not call a separate model API for this plugin.
Drafts, template settings, explanations and resume history remain in this client's storage until you clear that storage or the client removes it. They do not automatically synchronize across clients. OCR language data may be cached locally. Resume payloads have no application-side persistent retention period because the relay does not save them. An optional, explicitly labelled experimental upload mode sends a selected PDF directly to ChatGPT using its file API only when the client offers that API; OpenAI's file retention rules then apply.
You can inspect and edit extracted text before generation, correct the generated resume, rename saved resumes and download PDF copies. Clear this plugin's storage through your client's storage controls to remove local drafts and history. Removing local data or uninstalling the plugin does not delete existing ChatGPT messages, files or downloaded PDFs. Manage those copies separately. For privacy or infrastructure metadata requests contact micromvp@gmail.com; do not send us your resume unless needed for an explicitly requested support task. See the data deletion page.
Photo to Recipe: photos, recipes and local history
Photo to Recipe is published by Trustflow. It requires no separate account or login. Its purpose is to turn a food photo selected by you into a recipe with the current ChatGPT model. You choose the image and may enter servings, taste or dietary preferences. Do not upload identity documents, medical records, credentials or unrelated private images. Crop out faces and private details when they are not needed.
Data categories and purpose. The app decodes and re-encodes the selected photo into a resized image before transfer, rather than forwarding the original file and its original metadata. The relay receives that image, its media type, interface locale and a random unguessable image identifier. It temporarily holds the generated recipe, delivery state, creation time used for expiry and a local-save acknowledgement. Preferences are included in the request to ChatGPT; the relay does not maintain a dietary or personal profile. The current ChatGPT model interprets the image and writes the recipe. The relay does not call a separate model API.
Recipients. The chosen image and generated recipe pass through Trustflow's hosting infrastructure at iplugin.dev and OpenAI's ChatGPT service to complete your explicit request. Recipe search, when enabled and available, uses ChatGPT's search capability and its providers; following source links connects to those websites under their own policies. We do not send the image to a separate image recognition vendor, sell recipe histories or use them for advertising. Hosting and network providers necessarily handle connection metadata such as IP address, request time and route while serving HTTPS. The application does not log image bodies, recipe contents or user histories. Infrastructure security logs are separate from the transient recipe memory and follow the infrastructure retention described elsewhere in this policy.
Retention. Photos and recipe deliveries are held in bounded server memory for up to 30 minutes from upload, subject to earlier process restart or capacity limits. Expired entries are removed by periodic cleanup (every 15 seconds) or subsequent requests; they are inaccessible after their expiry. There is no server recipe database, account database or persistent image/history directory for this app. Server restarts remove pending deliveries. This transient relay is not a backup. The app saves completed history, including the resized source photo, as JSON in the current browser or ChatGPT client's local storage. Local records remain until you delete them, clear that client's app storage, or the client removes its storage. Pending delivery and acknowledgement data may also be kept locally for up to 30 minutes. Exported backup files remain wherever you choose to save them.
Your controls. Choose whether to upload a photo, edit the preferences before generation and turn recipe search off. Delete individual recipes using the app's delete control after confirming. Export history before clearing storage and import a backup to restore it on a device of your choice. Backups can include the source photo; keep and share them carefully. History does not automatically synchronize between devices or with our server. Removing the plugin or disconnecting its MCP connection may not erase local storage. We cannot remotely retrieve or delete your device-local history. Delete exported files and manage ChatGPT conversations separately; OpenAI retention and privacy controls apply to content held by ChatGPT. For help, contact micromvp@gmail.com without attaching private photos or backup files unless needed for an explicitly requested support task. See the data deletion instructions.
Photo to Pixel: local image processing and temporary export downloads
Photo to Pixel is operated by Trustflow. It has no account, login, payment, photo history or separate model API. Its purpose is deterministic conversion of a selected image into square pixels, numbered cell colors and downloadable files.
Data categories and purposes. The photo you choose, its filename, dimensions, decoded pixels, grid density and selected colors are processed in the app on your device. The MCP opening tool receives no image argument. When you click a download control inside ChatGPT, the app sends the generated PNG or HTML color document, its output filename, media type and byte length to iplugin.dev so the client can download a real HTTPS attachment. The HTML contains the sampled grid image and its color lookup code. These generated files may still reveal the subject of your photo, including personal information in the image or output filename. The original photo file and its original metadata are not uploaded. Export is optional; previewing or inspecting cells does not upload their image data. Standalone browser exports use local download links.
Recipients. Trustflow's hosting infrastructure handles the generated export when you request a download. Hosting and network providers necessarily handle technical connection information such as IP address, request timing and route while serving HTTPS. Export-route access and error logging are disabled in our Nginx configuration, and the application does not write export bodies, filenames or private tokens to request logs. We cannot guarantee every infrastructure provider's internal metadata retention. OpenAI handles ChatGPT and its app environment under its own policies. This plugin does not send photos or exports to a separate image vendor or model API, sell them, or use them for advertising.
Retention and access. Generated exports are stored only in bounded process memory for up to five minutes after upload, with no application database, disk directory or export backup. Links contain random 192-bit tokens and there is no export-listing endpoint. Anyone with a link can download the file until expiry; it is a bearer link, not an account-protected file. Expired links are inaccessible and return 404. Cleanup runs every second and on access; an already active transfer can finish before its memory is released. A process restart or service maintenance can remove a file sooner. The relay accepts at most 64 MiB per file, 96 MiB in aggregate including pending reservations, and 32 live exports. It may reject a download if capacity is reached. A manual language preference may remain in browser storage until cleared. The selected photo, pixels and table state remain in app memory for the current page; there is no persistent photo library.
Your controls. Choose whether to select a photo and whether to export it. Use the built-in synthetic sample instead of a personal photo when trying the app. Avoid selecting sensitive images or remove unrelated details first. Close or reload the app to release its in-memory photo; clear this app's browser storage to remove a saved language preference. Let a temporary export expire rather than share its link. There is no account-based deletion interface or persistent photo collection to retrieve. Delete downloaded PNG/HTML copies separately in your device's file manager and manage ChatGPT conversations separately. Uninstalling the plugin does not remove downloaded files. Contact micromvp@gmail.com for privacy questions or requests about information Trustflow actually holds; do not send private images or live export links unless needed for an explicitly requested support task. See Photo to Pixel deletion instructions.
Word Cloud Studio: remote rendering and temporary workspaces
Word Cloud Studio is operated by Trustflow to turn supplied text or word weights into editable typography. It has no account, login, paid checkout, advertising profile or separate model API. Unlike client-only image tools, its word_cloud renderer runs on iplugin.dev.
Data categories and purposes. We process source text or explicit word-frequency maps, ignored words, optional PNG masks, SVG silhouette paths, selected colors, layout dimensions and rendering options to create PNG and SVG previews. These inputs can contain personal information if you include it. The service returns the generated images, counts, private workspace capability and revision identifier needed to continue the same studio. These identifiers route requests; they are not an identity or telemetry profile. A manual interface-language preference is stored in the app’s browser storage. Your ChatGPT conversation and tool messages are separately handled by OpenAI.
Recipients. Trustflow hosting and network providers handle the rendering request and generated download. Network infrastructure necessarily processes connection information, such as IP address and request timing, while serving HTTPS. The application does not retain request logs, source text, masks, capabilities or image bodies in a database or disk directory; Nginx access and error logs for the route are disabled. Infrastructure providers may have their own technical metadata retention. Source text and masks are not sent to another model API, an image vendor, advertisers or data brokers.
Retention and access. A workspace keeps its latest preview, source and settings only in bounded server-process memory for up to 30 minutes from creation or rendering. Rendering again replaces the latest cached preview without server-side undo. Expired entries are swept every second and on access; memory can remain briefly in an active transfer or request. PNG download attachments are separately cached in memory for up to five minutes. A restart or capacity eviction can remove either sooner. There is no result-listing endpoint, account library, durable project history or application backup of these caches. Workspace capabilities and export links contain random 192-bit values. Anyone holding the corresponding capability can access that workspace, and anyone holding an export link can download it until expiry; these are bearer capabilities, not authenticated accounts. Avoid sharing them. The service limits workspaces to 64 and 96 MiB, and PNG exports to 32 and 64 MiB, with an 8 MiB maximum per export.
Your controls. Choose the text and masks you send, remove personal information before rendering, and use the included synthetic fixtures for evaluation. Export work you want to retain. Close or reload the studio to release its in-memory client draft; server copies expire automatically and may remain until their documented timeout. Clearing browser storage removes the saved language choice, not an active server cache. There is no identity-indexed server record or per-workspace deletion endpoint, so closing the page is not immediate remote deletion. Delete downloaded files separately in your file manager, and manage ChatGPT messages under OpenAI settings. Contact micromvp@gmail.com for questions or rights requests concerning information Trustflow actually holds; do not send sensitive source text or live capabilities unless needed for an explicitly requested support task. See Word Cloud Studio deletion information.