Privacy Policy
English translation. The legally binding version is the German one — switch with the DE toggle (top right).
This policy explains the processing of personal data when using meinKontext (“the Service”) — a central registry that delivers the operator's personal agent context (skills, base instructions, subagent definitions, integration manifests, fact memories and knowledge wikis) over the MCP protocol. Optional context plugins also process tasks, projects, calendar events, messages and their workflow history. The Service itself runs no generative AI. The one exception, currently a temporary test of search performance: a small local search model computes search vectors from the stored content inside the Service, within our own infrastructure. No data leaves our infrastructure and no external provider is involved; the search vectors are deleted together with their content.
Controller
Wach & Wegener UG (haftungsbeschränkt)
handelnd unter Yesterday
Kortumstr. 68
44787 Bochum
Deutschland
Represented by: Sidney Wach, Alexander Wegener
Email: hello@yesterday-ai.de
Overview of processing
Types of data processed: master data (e.g. name), contact data (email address), access and authentication data (password hashes, hashed API keys, OAuth client registrations), usage and metadata (access/audit logs, technical connection data), the content objects the user uploads, and product feedback and support conversations.
Categories of data subjects: registered users/operators, organization members provisioned via SSO/SCIM, invitation recipients, people requesting beta access and other persons whose data users include in context or plugin records.
Purposes: providing and operating the registry, handling beta-access and support requests, product improvement and error resolution, authentication and access control, security and abuse prevention, and traceability of access (audit).
Legal bases
- Performance of a contract / pre-contractual steps — Art. 6(1)(b) GDPR (providing the user account and the Service).
- Legitimate interests — Art. 6(1)(f) GDPR (secure, stable operation, audit logging, abuse prevention).
- Legal obligation — Art. 6(1)(c) GDPR, where statutory retention or disclosure duties apply.
The provisions of the German Federal Data Protection Act (BDSG) apply in addition.
Hosting and provision
The Service runs on infrastructure within the European Union. The following processors are used (Art. 28 GDPR):
- OVH (Object Storage, Germany region) — storage of the database backups (which also contain the stored content), including transaction logs used for recovery.
- Hetzner Online GmbH, Industriestr. 25, 91710 Gunzenhausen, Germany — server, container and self-hosted PostgreSQL infrastructure (Kubernetes cluster) running the Service and its active database.
- Brevo GmbH (formerly Sendinblue GmbH), Köpenicker Str. 126, 10179 Berlin, Germany — dispatch of transactional account emails (address verification, password reset, account-deletion confirmation). Processed for this purpose: email address, name and technical dispatch metadata; processing takes place within the EU.
Legal basis for the hosting: Art. 6(1)(f) GDPR.
User account and authentication
An account is required. Registration and sign-in process the email address, name and a password hash. Authentication uses OAuth 2.1 (with PKCE) or organization-bound API keys; API keys are stored only as a hash and the plaintext is shown exactly once at creation. A technically necessary session cookie is set for the login and consent pages. Legal basis: Art. 6(1)(b) GDPR.
Transactional emails
The Service sends account-related emails only — address verification, password reset and the confirmation of an account deletion. No newsletters, no advertising. Dispatch is handled by the processor Brevo GmbH (see “Hosting and provision”). Legal basis: Art. 6(1)(b) GDPR.
Invitations
A user may invite up to five people by email. For this, the recipient address entered by the inviting user is stored with the invitation code and used to send the invitation email (dispatch via the processor Brevo GmbH, see above); at the inviting user's request the same invitation can be sent again a limited number of times. The invited person is under no obligation; the address is not used for any other purpose. Legal basis: Art. 6(1)(f) GDPR (invitation at the user's initiative).
Beta-access requests
The “Request beta access” form (on the home page and at /zugang) processes your name, email address, how you work with AI today, the AI apps you select and an optional message, plus, where present, the page the request came from, the referring page and the campaign parameters (utm) of the address you opened. We use this information to handle your request and to contact you about access. The request is stored in our customer management system (Twenty CRM, open source, self-hosted on the Hetzner infrastructure named under “Hosting and provision”); no further service provider receives it. To prevent abuse, the IP address is counted in memory and discarded at the latest eleven minutes after the last request; it is not stored permanently. No cookie is set and nothing is stored on your device. Legal basis: Art. 6(1)(b) GDPR (pre-contractual request). Retention: see “Retention and deletion”.
Server log files
When the MCP endpoint is accessed, technical connection data is logged: the timestamp, the method called, the client's user agent and the identifier of the calling principal. IP addresses and access tokens are not stored in these logs. The logs serve operational security and error diagnosis and are kept short-term. Legal basis: Art. 6(1)(f) GDPR.
Audit logging
For traceability and security, one audit entry is stored per request (organization identifier, principal identifier, action performed, affected object identifiers, timestamp). Sent-invitation events also contain the recipient email address. Advisory content checks also store findings with excerpts of up to 60 characters from the uploaded content. These excerpts can contain personal data. The audit log is append-only; earlier entries are not rewritten when an account is deleted. Secret values must not be uploaded or included in logs. Legal basis: Art. 6(1)(f) GDPR.
Content objects
The content objects a user uploads (skills, base context, subagent definitions, integration manifests, wiki articles and memory facts) are stored versioned and access-restricted. If this content contains personal data, the uploading user or organization is responsible for its lawfulness. A publish-time check rejects secret-value-shaped strings (tokens, keys) — secrets belong in an integration manifest as an Infisical reference, not as a value.
Sharing and managed agents
Users can share selected objects, wikis and plugin content (such as task lists or calendars) with other accounts. Access follows current membership and permissions. Some plugins also restrict records to a fixed audience. Owner-managed agents act under separate credentials with explicitly granted content; they do not automatically receive all of the owner's context. Shared records can contain names, assignments, messages and attributed workflow history visible to the authorized recipients.
Product feedback and support
Through your signed-in connected app, its agent can submit reports directly to our support at your request and retrieve your own support conversation. No additional feedback permission or plugin is required. We process the report title and description, submitted diagnostic details, messages, your account and organization identifiers, timestamps and handling data such as status, category and assignment. Personal context, logs and diagnostics are not attached automatically. Please do not submit passwords, access keys or personal data unnecessary for handling the request.
Reports are stored separately from personal context in the operator's support area. Access is limited to explicitly authorized support staff and support apps they authorize, within their current permissions. Organization administrators do not automatically receive support access. Internal notes and handling records serve internal processing and do not appear in your conversation. Other users cannot access your reports.
We process support requests to perform the service contract (Art. 6(1)(b) GDPR). We process voluntary suggestions and the necessary internal handling on the basis of our legitimate interest in improving the Service and resolving errors (Art. 6(1)(f) GDPR). You have a right to object to processing on this basis. Revoking an app’s connection to your account also ends its feedback access. This does not delete reports already submitted.
Your account export includes your reports and the conversation visible to you. Internal notes and internal handling data are not included. Your statutory access and other data subject rights remain unaffected; contact hello@yesterday-ai.de to exercise them. For retention and erasure, see “Retention and deletion” below.
MCP clients (the user's own responsibility)
The Service delivers context content over the MCP protocol to the client the user chooses (e.g. Claude Code, Claude Desktop, ChatGPT, Langdock). Processing within the client itself — in particular forwarding inputs to AI providers — is the responsibility of the user and the respective client vendor and is not covered by this policy. The Service itself runs no generative AI; the local search model (see above) is an exception under temporary test and passes no content to third parties. When an operator-authorized support app processes data, the operator remains responsible for that processing; it is not the reporting user’s responsibility.
Retention and deletion
Personal data is deleted once the purpose of its processing no longer applies and no statutory retention obligations conflict. Account data and stored content in organizations in which you were the only member are removed from the active account when it is closed. Stored files are cleaned up afterwards; retries or a backlog can delay this cleanup. Older backups can still contain previously stored data and are not erased by account deletion. Accounts whose email address is never verified are selected for automatic deletion after 30 days. Server logs are kept short-term.
Deleting the reporting account also removes its feedback, conversation and internal notes from the support area. Revoking an app permission alone does not trigger this deletion. Automatic time-based deletion of closed reports is currently disabled; a fixed maximum retention period is therefore not currently enforced technically. Data may be retained only as long as needed to handle the request, resolve errors or meet a legal obligation. Erasure requests can be sent to hello@yesterday-ai.de. Older backups may still contain feedback. Recovery restores a selected database point; subsequent changes, deletions and permission revocations are not automatically included. The limitations below apply to remaining audit metadata.
The intended maximum retention for audit entries remains twelve months. The current implementation does not consistently meet this limit: audit entries are deleted in monthly groups using a cutoff twelve months in the past. This can retain an entry for almost thirteen months plus the maintenance delay. Failed maintenance can extend that period. Entries in the fallback storage are not covered by this automatic deletion and need separate cleanup. The current implementation therefore does not ensure a maximum audit retention period.
Rights of data subjects
You have the right to access (Art. 15), rectification (Art. 16), erasure (Art. 17), restriction of processing (Art. 18), data portability (Art. 20) and a right to object (Art. 21) to processing based on legitimate interests. You also have the right to lodge a complaint with a data protection supervisory authority (Art. 77 GDPR). A message to hello@yesterday-ai.de suffices.
Changes to this policy
We adjust this privacy policy when changes to the processing require it. The version published here applies.
Last updated: 2026-11-14