Skip to content
Official Legal DocumentVersion: 1.0 • October 28, 2026

MCP Usage and Data Transfer Terms

MCP connections to external AI clients, permission scopes, human approval, workspace data transfers and access revocation.

Service Provider: EMRE ALAY — CREATO AI • Karşıyaka / İzmir

Document version: v1.0-2026-10-28 — Date: October 28, 2026.

These terms supplement the Terms of Service and SaaS Agreement for Customers and authorized users connecting their CREATO AI workspace to an external AI client of their choice. Service provider: EMRE ALAY — CREATO AI. Contact: hello@creato.digital.

1 Nature of the feature and connection methods

Model Context Protocol (MCP) enables the User to connect ChatGPT (OpenAI), Claude (Anthropic), Cursor, Windsurf or another supported third-party AI client (External AI Client) to their own CREATO AI workspace. The connection provides access to reading, analysis and approved-action tools limited to the workspace and permission scope selected by the User.

Connections are established by granting authorization through the CREATO Portal login and consent screen using OAuth 2.1 and PKCE, or by generating a personal MCP access key (crk_live_...) in the Portal. The User checks the selected workspace, permission scope and connecting client during setup. A SHA-256 hash of the personal key is stored instead of its raw value.

An MCP connection is established on the User's active instruction. Before approving a connection or generating a personal key, the User must separately accept the following statement:

I have read the CREATO AI MCP Usage and Data Transfer Terms; I agree that workspace data within my selected permission scope will be shared with my connected third-party AI client.

This acceptance is a service instruction for the selected connection and scope. It is not blanket explicit consent on behalf of customers, leads or other individuals whose data is held in the workspace. Required notices, legal bases and conditions for processing and transferring personal data must be satisfied separately.

3 Permission scopes and human approval

  • Read Only — `READ_ONLY`: Accesses reading and reporting tools only; cannot create proposals or change data. Users with the Viewer (VIEWER) role are restricted to this scope.
  • Propose Only — `PROPOSE_ONLY`: Can read data and create pending (PENDING) change proposals. An authorized User must approve the proposal in CREATO Portal before it is applied.
  • Propose + Chat Confirm — `PROPOSE_AND_CONFIRM`: Can read data, create proposals and apply a proposal through a separate confirmation step after the User expressly approves it in chat. The two-stage Turn-Lock prevents proposal and execution from taking place as a single unapproved action.

No scope grants access to another workspace or permission to exceed the User's existing account privileges. Instructions from an External AI Client cannot remove these limits. The User reviews a proposal's content, target and effects before approval; general connection authorization does not replace approval of a specific change.

4 Data that may be transferred and access boundaries

Subject to the requested tool and selected scope, the following information from the relevant workspace may be transmitted to the External AI Client:

  • CRM and leads: Names, phone numbers, email addresses, status, follow-up dates, assigned team members and notes.
  • Tasks and team workflows: Task titles, descriptions, due dates, priorities and reminders.
  • Conversation intelligence: Conversation statistics, topic and intent distributions, knowledge gaps and objection summaries.
  • AI Employee information: Employee name, status, connected channel names, communication tone, task and campaign rules, frequently asked questions, enabled capability names and version history.
  • Knowledge base and support: Company knowledge document contents, revision history and CREATO support records.
  • Personal routines: Recurring reports and reminders configured by the User.

CREATO AI's core system instructions (systemPrompt) and internal model architecture; channel and integration credentials, encrypted tokens (accessToken, authValue), private API headers and user passwords are not exported through these tools. Data belonging to other brands or workspaces outside the User's authorization is excluded from access.

5 External clients and cross-border transfers

Tool responses are transmitted over HTTPS to the External AI Client connected by the User. The provider's servers or subprocessors may be located outside Türkiye. Before connecting, the User evaluates the selected provider's terms, processing regions, retention and deletion settings, and model-training preferences.

The External AI Client's retention, subsequent processing and any use of received data for model training depend on that provider's terms, the account and plan used, and applicable settings. CREATO AI does not operate that independent client or control its data practices. This distinction does not remove CREATO AI's obligations concerning its own technical service, fault, data security or statutory data protection duties.

The Customer, as data controller or authorized representative for workspace personal data, provides the required notices and legal basis. Cross-border transfers must meet the conditions of Article 9 of Turkish Law No. 6698 and applicable legislation, including GDPR where relevant. Checking the connection consent box does not by itself satisfy those conditions. Any required adequacy decision, appropriate safeguard or applicable exception must be assessed separately for the transfer concerned. The Turkish Data Protection Authority's cross-border transfer guidance explains these mechanisms.

6 Proposal accuracy and allocation of responsibility

An External AI Client may misinterpret instructions or generate incorrect content. The User is responsible for checking the accuracy, legal suitability and operational scope of a proposal before approval; the Customer is responsible for its source data, user authorizations and instructions. The User should not act solely on an AI suggestion without assessing the effects of the action being approved.

CREATO AI is responsible for technically enforcing the permission limits and approval steps within its control and providing the service with the agreed standard of care. Responsibility for an external client's independent activities is assessed under the relevant service terms and applicable law. The main agreement's provisions concerning fault, liability limits and mandatory law remain applicable; no clause automatically shifts responsibility for CREATO AI's own fault to the User.

7 Credential security and revocation

Personal MCP keys must be protected like passwords. They must not be shared with unauthorized parties or published in public code or messages. The User connects only trusted External AI Clients and limits the selected scope to their needs. If a key is exposed or suspicious activity occurs, the connection or key must be revoked immediately through CREATO Portal > Settings > MCP & AI Connection (Revoke).

The User may revoke a connection or personal key at any time through the same screen. Revocation stops new access through that connection or key; it does not automatically delete data previously transmitted to an external provider. That provider's deletion and retention controls must be used separately for such data.

8 Usage limits and intellectual property

MCP access is limited to permitted workspace operations. It must not be used for unauthorized data collection, scraping, attempts to access other workspaces, bypassing access controls or copying CREATO AI infrastructure and protected prompt architecture. The limit of 60 requests per minute per key and technical quotas must not be circumvented.

CREATO AI may restrict or suspend connections that threaten security or breach these rules in accordance with the main agreement's suspension provisions. Permission to use the feature does not transfer CREATO AI's intellectual property or third-party rights; rights granted by mandatory law remain unaffected.

When a connection or key is approved, the user identity, email address, IP address, browser information (User-Agent), timestamp, selected workspace (customerId), permission scope (scope), connecting client name and accepted document version are recorded in UserConsentLog. These details document the authorization's scope, time and accepted text.

The parties agree that properly maintained electronic records may be used as evidence in disputes under Article 193 of the Turkish Code of Civil Procedure (HMK) and applicable law. Rights to challenge the accuracy of records and submit contrary evidence remain protected; no provision is intended to make either party's ability to prove its case impossible or excessively difficult.

The Turkish and English texts relate to the same version, v1.0-2026-10-28. Updating this page does not change the version accepted in a historical consent record.

These terms are read together with the Terms of Service, Data Processing Agreement and Privacy Policy. They explain the data and permission scope of the MCP connection specifically enabled by the User; they do not create general authorization to transfer data to external AI services in other features. Mandatory use and data-processing restrictions of connected channels remain applicable.

For rights requests, connection security reports and legal inquiries: EMRE ALAY — CREATO AI, hello@creato.digital.

Have a Legal Question?

Contact us directly regarding contracts, data protection and rights requests.

hello@creato.digital