plugin.space / Flows

Flows by plugin.space

Privacy policy

How we handle information when you use Flows, connect an account or contact us.

Version 1.12 · 7 November 2026

Who is responsible

Michał Drożdżyński
ul. Bojowa 10c, 44-193 Knurów, Polska
Polish tax identification number (NIP): 9691629431
michal@drozdzynski.pl

Contact us at michal@drozdzynski.pl for privacy requests. We are the controller for account, service security, billing and support information. If an organisation uses Flows to store personal information in its own materials, it remains responsible for deciding why those materials are processed. Contact us before uploading content that requires a separate data processing agreement.

Local work and your AI host

Flow documents, screenshots, local discussions and the local library index are stored on your device. A document may also contain user-role definitions, permission conditions and evidence, user stories, test actions and assertions, quality-review hints and generated drafts. Without a connected account, opening a local flow does not upload it to our service. Connecting an account enables private cloud saving for opened and newly created flows, including their screenshots and source references, according to the account plan. Flows without a saved preference are uploaded automatically while their viewer or panel is active; later changed saves also upload automatically. Explicit per-flow opt-outs remain off, and preferences belonging to another account are not reused. You can turn them off at any time. An older sync setting without an account binding requires a new choice. You can also save manually or remove a cloud copy while retaining the local files. Local files and any copies you create remain under your control until you delete them.

Your AI coding tool may process selected code, images and conversation content under your account with that provider. Local storage in Flows does not mean the AI conversation runs offline. Flows does not call a separate paid model API on your behalf.

The local launcher reuses a compatible installed Node.js runtime or downloads a private runtime from nodejs.org and checks it against the checksum included in the plugin. That download sends ordinary network information, such as your IP address, to the runtime distributor; it does not send flow documents or screenshots. Setting FLOWS_OFFLINE=1 prevents this download and requires a compatible runtime to be available locally. Optional capture visits the application you select. Installing its browser runtime contacts npm and browser distribution services. Those services and the visited application receive ordinary network information, such as your IP address. Use authorised access and test data.

Connecting your assistant

If you connect the Flows MCP service, your selected AI host can read flows saved in the connected account and flows shared with it. Tool results can include diagram content, project names, source references, screenshots and the role, test and quality-review materials stored with a flow. The host also receives a stable account identifier and, when you grant email access, your verified email address. These results are processed under your relationship with that host. The hosted MCP panel does not upload local files or make purchases. The hosted connection reads permitted content. With a separate sharing authorisation, owners can manage the selected saved flows, grant or revoke email access, and enable or disable access for anyone with the link in the panel. Retained history for invited people requires a separate choice. This sends the selected flow identifiers, recipient email where applicable, history choice and link-access setting to our service. An older write permission does not grant sharing access. Hosted MCP cannot edit saved flow content, post comments or delete flows; manage those changes on the account website or through the trusted local plugin. Panel-only labels depend on the AI host enforcing them and are not independent proof of a human click. The hosted service checks sharing permission, connected account and ownership before changing access. The local MCP server runs with your computer permissions. Use a trusted host and approve only the access you intend. It contains no analytics and receives no access tokens in its page content.

You can disconnect the current connection in the panel or through the host’s connection settings. Your account settings list local and assistant connections and let you revoke one or all; you can also revoke all assistant connections on the Flows connection consent page. Revocation prevents further authorised access after the next access check; it does not remove content already retained by your AI host or copied into a conversation.

Information we receive

  • Accounts: your email, account identifier, display name if provided, verification status and session information, to authenticate you and provide requested access.
  • Cloud content: snapshots of flows you choose to save to a connected account, including automatic first saves and later updates while the connected viewer or panel is active, unless you disable automatic updates for that flow, their referenced screenshots, project names, source references and any information visible in them. Supported source fields also include user-role names and definitions, permission conditions and evidence, role assignments, user stories, acceptance criteria, test actions and assertions, and quality-review tasks and expectations. Cloud snapshots keep these source fields; viewers reconstruct the generated scenarios, use cases and quality suggestions from them. Local discussions are not included in a snapshot upload; comments posted to cloud discussions are sent separately.
  • Sharing: selected flow identifiers, teammate email addresses and history choices supplied by the owner, plus link-access settings and access tokens embedded in share URLs, to provide the requested access. If another user provides your email, that user is the source. Flows does not send marketing to those addresses.
  • Access checks: when the local viewer is connected, it asks our service to confirm the account’s current plan and local flow allowance, on connecting and at later online checks. These requests carry the connection’s authorisation and ordinary network information, not flow content.
  • Billing: Stripe customer and transaction references, subscription status and relevant accounting information. Stripe collects card details on its own hosted pages; we do not receive complete card numbers or security codes.
  • Support and security: messages you send us and operational records generated when you access the service, to respond to requests, investigate faults and protect accounts. To pause unusual bursts of automated writes, we keep one counter per account with the current short time window and the number of cloud writes in it; it is not a log of your activity.

Roles, generated materials and exports

Role audits describe the application you map and may expose internal role names, access rules, ownership conditions, source paths and unresolved security questions. These labels do not grant service permissions or create accounts for the people they describe. Test and quality suggestions are drafts derived from the selected document, not records of a test run, user study or automated decision about a person. Generating and previewing them does not send them to a separate evaluation or test-running service.

JSON, YAML and Markdown exports can include the selected flow’s role definitions, source evidence, test actions, expected outcomes and research proposals. A role filter narrows the exported scenario selection; it does not anonymise the content. When you copy or run an export in another service, that service receives the content you provide under your relationship with it. Review free text, source references and test data before saving, sharing or exporting.

Optional website analytics

With your consent, we use PostHog in its EU region to measure page visits, downloads, installation-guide interactions, successful sign-in and selected cloud-library actions. Events use a random browser identifier and limited device and campaign information. We do not send your email, verification codes, flow content, screenshots, comments or source code. We disable session recordings, automatic interaction capture and person profiles; project settings discard IP addresses. Requests pass through our website before reaching PostHog. The local plugin has no analytics.

Analytics is optional and uses consent as its legal basis (Article 6(1)(a)). We remember your choice for up to 180 days. After you allow analytics, a browser identifier is stored in localStorage. Choose Privacy settings to withdraw consent and clear that analytics identifier. The 180-day period applies to the consent choice, not to events already sent to PostHog. Withdrawing consent stops further collection in that browser; it does not itself delete previously collected events. Contact us to request deletion of analytics data. Your choice does not affect access to Flows. See PostHog’s privacy policy.

Why we use this information

We process account and service information to perform our agreement with you (GDPR Article 6(1)(b)). We process required accounting records to meet legal obligations (Article 6(1)(c)). Service security, responding to enquiries and managing access requested by an organisation rely on our legitimate interests in operating a secure service and handling those requests (Article 6(1)(f)), subject to your rights. We do not sell personal information, use customer content for advertising, or train our own AI models on it.

Who receives information

Our service uses Vercel for hosting, Neon for database and authentication services, Stripe for payments and PostHog for optional analytics. For purchases sold through Link, Stripe also acts as merchant of record and processes billing information for tax calculation, invoicing, fraud prevention and transaction support under its own privacy terms. We send your account email and a customer reference to Stripe when you open a purchase or billing session. They receive the information needed for those functions. Your selected teammates can see the flows and discussions you share with them, including future saved updates. If you enable Anyone with a link, anyone who obtains that URL can read the selected current flow documents and captures without signing in. This includes supported role definitions, permission evidence, source references, test hints, user stories and quality-review source fields in those documents. The public-link view excludes discussions and retained history. New project flows are not added to the shared selection automatically. Treat an enabled public link as a disclosure to an open audience: it can be forwarded, and disabling it cannot retrieve copies already made. Retained previous versions and their earlier screenshots are shared only if the owner separately enables history access, and are viewable while the owner has active Pro access. Authorised service operators may access information when necessary for support, security or a legal obligation.

Providers can process information outside Poland and the European Economic Area. The applicable provider terms and transfer safeguards govern those transfers; contact us for details relevant to your account. We do not claim that all data stays in the EU. Your independently selected AI host and applications visited during capture operate under their own terms.

Storage and retention

InformationRetention
Optional website analytics eventsOur current PostHog Cloud Free plan makes events from the past year available for analytics reports. This reporting window does not automatically delete older stored events. We have no configured age-based deletion schedule for analytics events, so there is no fixed automatic deletion deadline. Contact us to request deletion of events associated with your browser; we may ask for your random analytics identifier to locate them. Withdrawing consent stops new collection in that browser but does not erase earlier events. See PostHog’s event-retention documentation.
Local documents and settingsUntil you remove them from your device and any backups. We cannot remotely delete them. Deleting a local flow in Flows removes its document and local comments; screenshot files already captured and any saved cloud copy remain until you delete them.
Cloud account, saved flows and commentsAccount information remains while your account is open. Saved flows and comments remain until deleted or removed as part of an account-closure request. The same retention applies to role audits, user stories, test hints and quality-review source fields saved inside a flow. When an owner successfully deletes a saved flow, its snapshot, screenshots, comments, sharing records and saved version history are removed from active cloud storage. Local files and copies already downloaded by other people or retained by an AI host are not deleted. Contact us to request account closure. Ending a subscription does not delete saved flows automatically.
Saved flow version historyWith Pro, we keep up to 20 previous snapshots in addition to the current snapshot. Older versions are removed as new versions exceed that limit, and an owner with active Pro access can permanently remove individual versions. Previous snapshots can include earlier screenshots, source references, role evidence and test or quality-review source fields; changing the current flow does not immediately remove those earlier copies. When Pro access ends, retained versions are locked, not deleted: they stay stored until the flow is deleted or the account is closed, and become accessible again if Pro is renewed. Deleting the saved flow also removes its version history from active cloud storage.
Sharing recordsRecipient email addresses, history choices and selected flow identifiers are kept while the corresponding sharing configuration remains. Owners can remove invitations or change the selection. Disabling public link access prevents further anonymous use; project sharing can retain an invited-only configuration. Removing a flow prevents it being retrieved through a project selection. Contact us to include remaining project-sharing configuration in an account-closure request.
Cloud write counterOne record per account, holding only the current short time window and its count. It is overwritten as you continue to save and is not kept as a history.
Local viewer connectionAuthorisation codes expire after 5 minutes and can be used once. Connections expire within 24 hours or when revoked. Expired database records are cleaned during subsequent connection requests.
MCP assistant connectionAuthorization codes expire after 5 minutes. Access tokens expire within 15 minutes; rotating refresh tokens can keep a connection active for up to 30 days unless revoked. Tokens are stored as hashes. Expired grants and sessions are removed during subsequent authorization requests. Client registration metadata remains while the client is registered; contact us to request its removal.
Support correspondenceUntil the request is resolved and for up to 24 months afterwards, unless a specific dispute or legal obligation requires longer retention.
Payment and accounting recordsFor the applicable statutory accounting and tax retention period. Records retained solely for that purpose are not used to keep a closed account active.
Application runtime logsVercel retains runtime logs for up to 30 days under our current Observability Plus configuration. These logs contain technical request and error information used to operate and protect the service. We have no configured log drains forwarding a separate copy to another logging service.
Database recovery historyOur Neon database has a rolling 6-hour point-in-time recovery window. Data changed or deleted from active storage can remain recoverable within that window until it ages out. No additional database snapshots or snapshot schedule are configured. This recovery window does not determine the retention of saved flow versions, runtime logs or payment records described separately above.

Cookies and browser settings

Essential authentication cookies keep you signed in on the account website. The local viewer uses an HttpOnly connection cookie, with the actual access token held in its Node process; restarting that process ends the cloud connection. A separate local cache stores the connected account identifier, service origin, local flow allowance and verification/expiry times, without an access token. It supports offline local access for up to 7 days after the last online check, never beyond the account’s confirmed access: the paid period for a subscription, or the next required online check for access granted without a subscription. Disconnecting in the local viewer clears it; remote revocation is observed on the next online check. The viewer also stores theme, graph-view and local comment-name preferences in browser storage. We do not include advertising trackers in Flows. Browser or provider settings may impose additional storage limits.

Your choices and rights

Write to michal@drozdzynski.pl to request access, correction, deletion, restriction, portability or to object where applicable. Include your account email and the request; do not send passwords, codes or identification documents. We verify identity proportionately and respond within the applicable legal deadline, normally one month for GDPR requests, explaining any permitted extension.

The viewer exports flow JSON and draft test or research materials as JSON, YAML or Markdown as applicable. Those exports do not include every screenshot, discussion or account record; contact us for a complete account-data request or closure. You can revoke invitations, disable public link access and disconnect local access in Flows. Revocation cannot retrieve copies other users already downloaded. You can complain to the Polish data protection authority, UODO, or your competent supervisory authority.

Data incidents

Report suspected unauthorised access or disclosure to michal@drozdzynski.pl with the affected account, approximate time and a short description. Do not email passwords, access tokens or unredacted screenshots. We investigate, contain incidents and document our assessment. As a processor, we notify the affected controller without undue delay after becoming aware of a personal data breach. As a controller, we notify the competent supervisory authority without undue delay and, where feasible, within 72 hours of awareness, unless the breach is unlikely to result in a risk to people’s rights and freedoms. We inform affected people without undue delay where the breach is likely to result in a high risk, subject to the exceptions in applicable law.

Security and changes

Do not upload credentials, payment-card data, government identifiers, protected health information or content you are not authorised to process. Cloud uploads keep only supported flow fields and remove URL credentials, query parameters and fragments from structured source and capture references. This is not anonymisation: free text, URL paths and image pixels can still contain secrets or personal information. Inspect every capture and use test data. The service uses authentication and access checks but is not end-to-end encrypted storage. No service can promise perfect security. We publish changes here with a new version date and provide notice of material changes where required.