New with Salesforce: Why 80% of peak carts still didn’t convert. Get the free guide →

Noibu MCP Terms of Service

Effective date: November 4th 2026

These Noibu MCP Terms of Service (the “MCP Terms”) govern Customer’s use of Noibu MCP features made available by Noibu Technologies Inc. (“Noibu”). These MCP Terms are incorporated into and form part of the master agreement between Customer and Noibu under which Customer accesses Noibu’s services (the “Agreement”). Capitalized terms used but not defined herein have the meanings given in the Agreement.

Key Principles

Noibu MCP enables Customer’s connected AI Model to (i) access Noibu Data, (ii) perform Write Actions on records within Customer’s Noibu account, (iii) perform Connected System Actions in third-party systems Customer has connected to Noibu (such as Customer’s code repository or ecommerce platform), and (iv) start Noibu Agent Tasks, in which Noibu-operated AI agents generate proposed code changes, experiments, audits or other work product for Customer’s review. Other than serving experiment variations that Customer has started (see Section 4.4), Noibu MCP never publishes a theme, merges code, or deploys to Customer’s production environment; those steps remain with Customer.

Authorization is established via an OAuth 2.0 flow into Customer’s Noibu account, facilitated by an Auth Provider that performs the authorization handshake only and does not store tokens or data. Connected System Actions use the integration credentials Customer has separately granted to Noibu for that system. Noibu Data is served by Noibu’s own systems directly to Customer’s AI Model session. Noibu Data is Customer Data and may include End User Personal Data captured in session recordings, heatmaps, and other behavioral telemetry. Noibu does not use Customer Data or Noibu Data to train AI models. Once Noibu Data is received by Customer’s AI Model, it is subject to the AI Provider’s own privacy, security, training, and retention practices, which Customer is responsible for reviewing.

1. Definitions

“Noibu MCP” means Noibu’s Model Context Protocol server, Noibu Skills, and associated features that enable Customer’s connected AI Model to (i) access Noibu Data, (ii) perform Write Actions on records within Customer’s Noibu account, (iii) perform Connected System Actions, and (iv) start Noibu Agent Tasks (each as further described in Section 4).

“Noibu Data” means data in Customer’s Noibu account, including but not limited to session recordings, error data, performance and analytics data, heatmaps, user behavior data, dashboards, issue and ticket data, account configurations, releases and deploy events, experiment configurations and results, custom event data, platform and server-side logs Customer has routed to Noibu, HTTP request and response payloads captured in sessions, accessibility audit results, Business Context, Call Metadata, Connected System Data that Noibu retrieves or caches, and any End User Personal Data captured therein. Noibu Data forms part of Customer Data under the Agreement.

“AI Model” means any third-party AI model, assistant, agent, coding tool, development environment extension, or other Model Context Protocol-compatible client that is procured, configured, and operated separately by Customer (the provider of any such AI Model, an “AI Provider”). The AI Model is not a Noibu product, and Noibu does not endorse, certify, or guarantee compatibility with any particular AI Model or AI Provider. Except for Noibu Agent Tasks (see Section 4.3), Noibu MCP acts as a passive connector and skill layer that executes tool calls directed to it by the AI Model. Customer is solely responsible for selecting the AI Model and for reviewing and complying with the applicable AI Provider’s terms of service, privacy policy, and data processing terms prior to connecting that AI Provider to Noibu MCP.

“Auth Provider” means the third-party authentication service Noibu uses to facilitate OAuth authorization between Customer’s AI Model and Customer’s Noibu account (currently Stytch). The Auth Provider handles the OAuth authorization flow only and does not store OAuth tokens or Noibu Data.

“Write Action” means an action executed by Noibu MCP, in response to a tool call from Customer’s AI Model, that creates, modifies or deletes a record or setting within Customer’s Noibu account (for example, changing the status, priority, owner, tags or notes of an issue or ticket; logging or editing a release; creating, editing, starting, stopping or concluding an experiment; saving Business Context; or starting an accessibility audit or other Noibu scan of Customer’s website). Some Write Actions (for example, starting an experiment) change what End Users see on Customer’s website, as described in Section 4.4. Actions in systems outside Customer’s Noibu account are Connected System Actions, not Write Actions.

“Connected System” means a third-party system that Customer has connected to its Noibu account through a Noibu integration and for which Customer has granted Noibu credentials, including a source code hosting service (for example GitHub or Bitbucket), an ecommerce platform (for example Shopify), and a project or ticket management tool (for example Linear or Jira).

“Connected System Action” means an action executed by Noibu MCP or a Noibu Agent, in response to a tool call from Customer’s AI Model or as part of a Noibu Agent Task, that reads from or writes to a Connected System using the credentials Customer granted to Noibu (for example, reading repository or theme files; creating a branch and opening a pull request containing proposed code; creating, editing, previewing and deleting unpublished preview copies of Customer’s storefront theme; and creating or updating tickets). Connected System Actions do not include merging a pull request, publishing a theme, deploying code, or otherwise changing what is live on Customer’s production website, each of which must be performed by Customer in the Connected System itself.

“Connected System Data” means data Noibu retrieves from a Connected System in the course of a Connected System Action, such as source code, theme files, and ticket content.

“Noibu Agent” means an AI-powered capability operated by Noibu on Noibu’s infrastructure (for example Noibu Autofix, Noibu’s experiment builder, and Noibu’s accessibility auditor). A “Noibu Agent Task” is a unit of work performed by a Noibu Agent after Customer’s AI Model or Customer starts it through Noibu MCP or the Noibu console, such as diagnosing an issue and opening a pull request containing a proposed fix.

“Noibu Skill” means prewritten instructions, workflows, prompts, rules or plugins that Noibu publishes for use with Customer’s AI Model and that direct the AI Model to make tool calls to Noibu MCP.

“Business Context” means information about Customer’s business (such as goals, site structure, deployment process and team setup) that Customer or its AI Model saves to Customer’s Noibu account for use by Noibu Skills and Customer’s AI Model.

“Call Metadata” means the operational metadata Noibu MCP records about each tool call, including the tool invoked, its parameters, the AI Model’s stated rationale for the call, the Noibu Skill (if any) and skill step that produced the call, timestamps, and the Customer user and account associated with the authorization.

“End User” has the meaning given in the Agreement and generally refers to individuals who interact with Customer’s websites or applications and whose activity may be captured in Noibu Data.

2. Authorization Model

Connections between Customer’s AI Model and Customer’s Noibu account are established exclusively via OAuth 2.0 authorization. The following applies to all such connections:

  • Customer initiates the connection by adding Noibu MCP to its AI Model and is redirected to a Noibu-hosted OAuth authorization screen, which is facilitated by the Auth Provider.
  • Customer grants the connection by authenticating into Customer’s Noibu account through the authorization screen. The scope of access is limited to the Noibu Data, Write Actions, Connected System Actions and Noibu Agent Tasks described in these MCP Terms and available to that user under Customer’s Noibu account permissions.
  • OAuth Tokens issued through this flow are not persistently stored by the Auth Provider, and Noibu does not receive, see, or store Customer’s Noibu account login credentials.
  • OAuth Tokens are used solely to authenticate tool calls from Customer’s AI Model to Noibu MCP and authorize access to Noibu Data, Write Actions and Noibu Agent Tasks within Customer’s account.
  • Connected System Actions do not use the OAuth Token issued to the AI Model. They use the integration credentials (for example an installed Shopify app or GitHub app) that Customer separately configured in its Noibu account, and are limited to the permissions Customer granted to that integration. Noibu MCP cannot act in a Connected System that Customer has not connected.
  • Customer may revoke the authorization at any time through Customer’s Noibu account settings or by disconnecting the AI Model integration. Revocation immediately terminates Noibu MCP’s ability to respond to tool calls from that AI Model. Customer may separately disconnect any Connected System through its Noibu account settings or by uninstalling the Noibu integration in that system, which disables Connected System Actions for that system.

Because authorization is granted directly by Customer through a Noibu-hosted OAuth flow, and Connected System access is granted directly by Customer through each integration, the scope of access available to Noibu MCP is controlled by Customer and may be revoked at any time.

3. Noibu Data Access

3.1 Read Access to Noibu Data. Noibu MCP enables Customer’s connected AI Model to retrieve Noibu Data. Noibu Data is served by Noibu’s own systems directly to Customer’s AI Model session in response to AI Model tool calls. Categories of Noibu Data that may be accessible include session recordings and replay data, error data, performance and diagnostic metrics, heatmaps and user behavior data, analytics and dashboards, issue and ticket data, account configurations and settings, releases, experiment configurations and results, custom events, platform logs, HTTP payloads captured in sessions, accessibility audit results, Business Context, and Connected System Data.

3.2 Customer Data Characterization. Noibu Data is Customer Data under the Agreement. By enabling Noibu MCP and connecting an AI Model, Customer instructs Noibu to make Noibu Data available to Customer’s chosen AI Model on Customer’s behalf. Such transmission is at Customer’s direction and is not a disclosure of Customer Data by Noibu to a third party for purposes of the Agreement. Once Noibu Data is received by Customer’s AI Model, it is processed by the AI Provider subject to the AI Provider’s own terms of service, privacy policy, training, retention, and security practices, and Noibu has no control over and no liability for the AI Provider’s handling of that data.

3.3 End User Personal Data in Noibu Data. Customer acknowledges that Noibu Data, including session recordings, heatmaps, HTTP payloads, platform logs, and related behavioral telemetry, may contain Personal Data relating to Customer’s End Users (such as form inputs, click trails, navigation paths, or other behavioral signals), including information that may not have been masked under Customer’s session-capture configuration. When Customer enables Noibu MCP and routes Noibu Data to its AI Model, such End User Personal Data may be transmitted to and processed by the AI Provider. Customer is solely responsible for: (a) ensuring its session-capture and data-masking configurations are appropriate for transmitting Noibu Data to its AI Provider; (b) ensuring its own privacy notices, terms of service, and End User consents adequately disclose and authorize such onward processing by the AI Provider; (c) determining whether the AI Provider’s training, retention, and security practices are acceptable for the categories of data being transmitted; (d) complying with all applicable data protection laws in connection with such transmission; and (e) any cross-border transfer of Noibu Data resulting from Customer’s choice of AI Provider.

3.4 Connected System Data. When Customer’s AI Model directs a Connected System Action that reads from a Connected System, the resulting Connected System Data (for example theme files or source code) is retrieved by Noibu using Customer’s integration credentials and returned to Customer’s AI Model session or, for Noibu Agent Tasks, processed by the Noibu Agent as described in Section 4.3. Connected System Data is Customer Data. Customer is responsible for ensuring it has the right to grant Noibu and its AI Provider access to the Connected System and its contents, including any third-party or licensed code it contains, and for ensuring the Connected System integration is scoped to the repositories, stores and projects Customer intends.

3.5 Call Metadata. Noibu records Call Metadata for each tool call made to Noibu MCP. Noibu uses Call Metadata to operate, secure, audit and troubleshoot Noibu MCP, to enforce usage limits and these MCP Terms, and, in aggregated or de-identified form, to understand how Noibu MCP and Noibu Skills are used so that Noibu can improve them. The AI Model’s stated rationale may paraphrase instructions Customer’s users gave to the AI Model; Customer should instruct its users not to include End User Personal Data or confidential information in prompts beyond what is needed for the task. Call Metadata is retained for [TBD — retention period] and is not used to train AI models.

4. Write Actions, Connected System Actions and Noibu Agent Tasks

4.1 Scope of Write Actions. In addition to read access to Noibu Data, Noibu MCP supports Write Actions that allow Customer’s AI Model to create, modify or delete records and settings within Customer’s Noibu account. Write Actions currently include: (a) issue and ticket lifecycle and metadata changes, such as closing, reopening or changing status, assigning or unassigning owners, adding or modifying tags, labels or priorities, and adding comments or notes; (b) logging and editing releases and deploy events; (c) creating, editing and advancing the lifecycle of experiments (see Section 4.4); (d) saving and updating Business Context; and (e) starting Noibu scans and audits of Customer’s website, such as an accessibility audit, which causes Noibu’s crawler to visit Customer’s pages. Noibu may add further Write Actions of the same general character in accordance with Section 5.2.

4.2 Connected System Actions. Where Customer has connected a Connected System to Noibu, Noibu MCP and Noibu Agents may perform Connected System Actions in that system on Customer’s behalf. Connected System Actions are designed so that Customer retains the final step: Noibu MCP may create branches, commits and pull requests, but does not merge them; it may create, edit, preview and delete unpublished preview themes, but does not publish a theme to Customer’s live storefront; and it may create or update tickets, but does not alter Customer’s production systems. Customer acknowledges that Connected System Actions nonetheless create, modify and in some cases delete artifacts inside Customer’s own systems, are visible to other users of those systems, are recorded in those systems’ histories and audit logs, and may trigger those systems’ own automations, notifications, continuous-integration pipelines or usage charges. Noibu is not responsible for the configuration or behavior of Customer’s Connected Systems.

4.3 Noibu Agent Tasks. Certain Noibu MCP tools start Noibu Agent Tasks. Once started, a Noibu Agent runs on Noibu’s infrastructure, may read Noibu Data and Connected System Data relevant to the task, uses AI models provided by Noibu’s AI sub-processors (listed on Noibu’s Sub-Processors page) to analyze that data and generate work product, and delivers that work product through Connected System Actions (for example a pull request or preview theme) or within Customer’s Noibu account. For Noibu Agent Tasks, Noibu is not a passive connector: the content of the generated work product is produced by Noibu’s systems rather than by Customer’s AI Model. All Noibu Agent work product, including code, theme changes, experiment variations and audit findings, is a proposal provided “as is” for Customer’s review. It may contain errors, may not follow Customer’s coding or design standards, and may have unintended effects if deployed without review. Customer is solely responsible for reviewing, testing and approving Noibu Agent work product before merging, publishing or otherwise deploying it. Noibu Agent Tasks may consume credits or be subject to usage limits set out in the Agreement or Customer’s order form.

4.4 Experiments and Live Traffic. Noibu MCP allows Customer’s AI Model to create experiments (A/B tests) and advance them through their lifecycle. Starting an experiment causes Noibu’s script on Customer’s website to assign End Users to variations and to serve variation code to them in real time; stopping an experiment ends that assignment permanently and cannot be undone. Experiments therefore change what End Users see on Customer’s live website even though the experiment record itself lives in Customer’s Noibu account. Customer is solely responsible for reviewing variation code and targeting before an experiment is started, for ensuring experiments comply with Customer’s own policies and applicable law (including consumer protection, accessibility and privacy law), and for the effect of experiments on End Users and on Customer’s business.

4.5 Initiation of Actions. Write Actions, Connected System Actions and Noibu Agent Tasks may be initiated in three ways: (a) Customer-Directed Actions, where Customer or Customer’s authorized users provide instructions directly to the AI Model, which then directs Noibu MCP to execute the action; (b) Skill-Directed Actions, where Customer activates a Noibu Skill, or a skill that Customer has subsequently modified or configured, which provides instructions to the AI Model on Customer’s behalf; and (c) Scheduled or Automated Actions, where Customer configures its AI Provider, a Noibu Skill or another automation to run on a schedule or in response to a trigger without a human prompt for each run. Scheduled or Automated Actions are deemed Customer-Directed Actions for every run. Where Customer has modified, customized, or extended a Noibu Skill after initial setup, Noibu is not responsible for the instructions contained in the modified skill, and full accountability for the actions taken by the AI Model in response to that modified skill rests with Customer. In all cases, other than the generation of Noibu Agent work product under Section 4.3, Noibu MCP acts as a passive executor of tool calls it receives from the AI Model. Noibu does not control, predict, or guarantee the specific actions the AI Model will take in response to any instruction, whether Customer-provided, Noibu-provided, or Customer-modified.

4.6 Customer Responsibility for AI Model Behavior. Customer is solely responsible for supervising the AI Model’s use of Noibu MCP, including reviewing Write Actions and Connected System Actions before confirming them where the AI Provider’s interface supports confirmation, reviewing and testing pull requests, preview themes, experiment variations and other work product before merging, publishing or starting them, limiting the Noibu account permissions and Connected System permissions available to users who connect an AI Model, keeping its AI Provider sessions and automations secure, and ensuring that actions executed by the AI Model are appropriate for Customer’s workflows. Because Noibu MCP acts only on tool calls it receives from the AI Model, and the AI Model acts on Customer’s instructions or on Customer-activated skills, the chain of authorization runs from Customer through the AI Model to Noibu MCP and, for Connected System Actions, onward through the integration credentials Customer granted.

4.7 Reversibility. Write Actions performed on issues and tickets are generally reversible through Customer’s Noibu account (for example, a closed issue can be reopened). However, certain operations may have downstream effects (such as triggering notifications to assignees or updating reporting metrics) that are not automatically reversible. Some Write Actions and Connected System Actions are not reversible at all. For example, stopping an experiment is permanent; deleting a preview theme permanently removes it from Customer’s store; a Noibu Agent Task that has started cannot be un-run and its credits are consumed; a crawler visit to Customer’s website cannot be recalled; and branches, commits, pull requests and tickets created in a Connected System remain in that system’s history even after they are closed. Noibu MCP does not delete published themes, production branches or Customer’s live site content. Customer is responsible for any consequences arising from Write Actions, Connected System Actions and Noibu Agent Tasks executed at Customer’s instruction or through Customer-activated skills or automations.

4.8 No Liability for AI-Initiated Actions. All Write Actions, Connected System Actions and Noibu Agent Tasks executed through Noibu MCP are initiated by Customer’s use of the AI Model or by Customer’s automations and are deemed authorized by Customer. Noibu does not independently initiate Write Actions, Connected System Actions or Noibu Agent Tasks and shall have no liability for such actions executed in response to AI Model tool calls authorized by Customer, nor for Customer’s decision to merge, publish, deploy or start any work product generated through Noibu MCP. To the maximum extent permitted by applicable law, Noibu Agent work product is provided without warranty of any kind, and Noibu’s liability in connection with Noibu MCP is subject to the limitations and exclusions of liability in the Agreement.

4.9 Actions Outside Noibu MCP. Certain AI Models, including AI coding agents and development tools, may be capable of taking actions outside Customer’s Noibu account and outside Noibu MCP (for example, modifying source code, creating pull requests, or deploying changes using the AI Model’s own access to Customer’s systems) based on Noibu Data retrieved through Noibu MCP. Such actions are performed by the AI Model under Customer’s configuration and control, are not Write Actions or Connected System Actions, and are not performed by Noibu. Customer is solely responsible for reviewing, testing, and approving any such actions, and Noibu shall have no liability for any action taken by an AI Model outside Noibu MCP.

5. Noibu MCP Availability and Changes

5.1 Third-Party Dependency. Noibu MCP functionality depends on Customer’s AI Provider, the Auth Provider, Customer’s Connected Systems and their providers, Noibu’s AI sub-processors, and the underlying MCP protocol, each of which is outside Noibu’s control. Noibu does not warrant that Noibu MCP will remain available, uninterrupted, or functional. The Auth Provider, the AI Provider, a Connected System provider, or the MCP protocol itself may experience outages, modifications, or changes (including changes to their APIs, rate limits or permission models) that interrupt Noibu MCP functionality.

5.2 Changes to Noibu MCP. Noibu may modify the scope of Noibu Data accessible via Noibu MCP, the set of Write Actions, Connected System Actions and Noibu Agent Tasks supported, the Noibu Skills it publishes, or the underlying technical implementation, with reasonable notice to Customer. Noibu may change the Auth Provider or its AI sub-processors, or use additional sub-processors, with reasonable notice to Customer and by updating its Sub-Processors page.

5.3 Customer Credentials. Customer is solely responsible for managing access to its Noibu account and to its Connected Systems, and for any consequences resulting from unauthorized access to the AI Model session connected to Customer’s Noibu account or to any automation Customer has configured to use Noibu MCP.

6. Data, Privacy, Security

6.1 No AI Training Use. Noibu does not use Customer Data, Noibu Data, Connected System Data, Call Metadata, or any data transmitted through Noibu MCP to train Noibu’s or any third party’s AI models. Noibu’s agreements with the AI sub-processors used for Noibu Agent Tasks restrict those sub-processors to processing Customer Data solely to provide services to Noibu and prohibit use of Customer Data to train their models.

6.2 Data Processing Addendum. To the extent Noibu processes personal data contained in Noibu Data or Connected System Data in its capacity as a data processor on behalf of Customer, such processing is governed by the Data Processing Addendum (“DPA”) attached to or incorporated into the Agreement. The Auth Provider acts as a sub-processor with respect to OAuth authorization into Customer’s Noibu account, Noibu’s AI sub-processors act as sub-processors with respect to Noibu Agent Tasks, and Customer’s acceptance of these MCP Terms constitutes Customer’s authorization of each of them as a sub-processor for those purposes. Customer’s AI Provider and Customer’s Connected System providers are engaged by Customer, not by Noibu, and are not Noibu sub-processors.

6.3 Security. Noibu maintains appropriate technical and organizational measures to protect the components of Noibu MCP that Noibu operates, consistent with the security standards set out in the Agreement. Connected System credentials granted to Noibu are stored encrypted and used only for the Connected System Actions and Noibu Agent Tasks described in these MCP Terms. Customer should grant each Connected System integration the minimum permissions necessary. Customer acknowledges that OAuth authorization is facilitated by the Auth Provider, and that the security of the authorization handshake is governed by the Auth Provider’s own security practices and the terms agreed between Noibu and the Auth Provider. Noibu performs reasonable due diligence on the security posture of its Auth Provider and AI sub-processors but does not warrant the security practices of any third party. Once Noibu Data is delivered to Customer’s AI Model, the security and handling of that data is subject to the AI Provider’s own security and data-handling practices, which Customer is responsible for reviewing. Once work product is delivered to a Connected System, its security is subject to that system’s and Customer’s own controls.

7. Termination of Noibu MCP Access

Customer may disable Noibu MCP access at any time by revoking the OAuth authorization through Customer’s Noibu account settings or by disabling the Noibu MCP integration in Customer’s AI Provider. Customer may disable Connected System Actions for any Connected System by disconnecting that integration in its Noibu account or in the Connected System. Disabling Noibu MCP does not reverse Write Actions, Connected System Actions or Noibu Agent Tasks already executed and does not stop an experiment that is already running, which Customer may stop in its Noibu account. Noibu may suspend or terminate Customer’s access to Noibu MCP for material breach of these MCP Terms, the Agreement, or applicable law. Termination of Noibu MCP access does not affect Customer’s underlying Noibu account or rights to Noibu Data, which continue to be governed by the Agreement.

8. General

8.1 Conflict with Agreement. In the event of a conflict between these MCP Terms and the Agreement with respect to Noibu MCP features, these MCP Terms shall govern solely with respect to such features.

8.2 Changes to These MCP Terms. Noibu may update these MCP Terms from time to time to reflect changes in product functionality, sub-processors, applicable law, or industry practice. Noibu will provide reasonable notice of material changes by email or in-product notification. Customer’s continued use of Noibu MCP features following such notice constitutes acceptance of the updated MCP Terms.

8.3 Contact. Questions about these MCP Terms may be directed to privacy@noibu.com or to Noibu’s Privacy Officer at the address set out in the Noibu Privacy Policy.

9. Governing Law

These MCP Terms are governed by the same governing law and jurisdiction provisions as the Software Services Agreement ("MSA")