Model Context Protocol

Connect your agents and AI tools to Orange Logic with MCP

Model Context Protocol (MCP) is an open standard that lets AI applications, such as Claude or Cursor, connect directly to external systems. Users can ask a connected AI application to retrieve asset information and update asset details without switching applications.

For example, a campaign coordinator can find a product photo, review its metadata, and ask the AI application to update a field or add a comment. Actions are performed on behalf of the signed-in user and are subject to their permissions.

Orange Logic runs an MCP server at /mcp.

📘

Note

MCP is available and activated by default beginning in Orange Logic Rome.

MCP tools

Depending on the user’s permissions and your environment’s tools, users can ask a connected AI application to perform the following tasks:

  • Find assets and retrieve records: Search for assets.
  • Review and update fields: Find out which fields you can edit, look up available values in a dropdown field, and update supported fields on one or more records.
  • Post a comment: Add a text comment to a record using natural language or markdown. Comments made through MCP cannot include attachments.
  • Relate tasks: Create a relationship between records.
  • Change a record’s status: View available statuses and update a task's status. Statuses are included in the Creative Collaboration package.
  • Manage participants: Add or remove participants in a role. Roles are included in the Creative Collaboration package.

For example, a project manager could ask an AI application to find a product launch project, update its status, assign a participant to the product marketing role, and post a comment with a progress update.

Available tools and actions depend on your environment's configuration, active modules, and your access level. The application can only perform supported actions that you are permitted to perform.

🎓

Insight: Relationships

All objects can relate to other objects. For example, consider an image that is in a folder. This relationship is a “Parent folder -primary content” relationship, and it is the primary way assets are organized in the Orange Logic platform.
Orange Logic has many built-in relationships. For example, the “Supported asset-supporting documents” relationship can be used to establish relationships between an asset and its supporting files, such as a retail product and its legal documents.

Each relationship has a parent and child.

  • In the parent folder-primary content relationship, the folder is the “parent” and the image is the “child."
  • In the supported asset-supporting documents relationship, the asset is the “parent,” and the supporting document is the “child.”

Each relationship has rules that determine the following:

  • Whether a child asset inherits properties from the parent asset.
  • Whether the relationship changes the child asset’s location.

MCP search

Orange Logic’s MCP allows users to search for Orange Logic assets in natural language within an AI tool. For example, suppose a marketing coordinator is drafting a campaign brief in Claude. Instead of opening Orange Logic to find the approved hero shot, they ask Claude to find it. Claude returns a link to the preview, the photo’s unique ID, and other information. The coordinator can use that result in their next task without leaving the AI application. They might add the preview to a mock-up, attach the asset's details to a project ticket, or pass its link to another connected tool. Orange Logic supplies the asset and enforces the governance, while the user and AI application orchestrate the asset through the rest of the project.

The table below lists additional examples to use Orange Logic’s MCP search.

StakeholderNatural language requestMCP server action
Brand manager"Show me the approved hero images from the spring campaign."Searches within the spring campaign for assets this user is allowed to access and returns each match.
Campaign coordinator"Find the latest product photos we can use on social media."Runs a natural-language search and sorts results by most recently modified because the request implies recency.
Photo editor"Which of the luxury chaise lounge photos are high-res enough for print?"Returns each luxury chaise lounge product photo with its dimensions and file size so the editor can evaluate the resolution and size without checking each photo Orange Logic.
Rights reviewer"Can we still use the Mango Home lifestyle set in North America?"Returns the rights summary in the asset's metadata (for example, "Perpetuity, North America, Web").
Developer or evaluator"What can I search for in this environment?"Reports the searchable fields, sort options, and whether AI search is available for that environment.

How search works

MCP searches use Orange Logic's search engine, security, governance, and rights. When users search Orange Logic from an AI application, the results match exactly what they would see in the Orange Logic interface.

Users can search for assets based on any of the following metadata:

  • Title
  • Unique identifier
  • Object type, such as a standard image, marketing campaign, product video, etc.
  • The name of the parent object, such as a folder, collection, campaign, project, or product
  • Aspect ratio
  • File size
  • Date created
  • Date last modified
  • Name of the record’s container (folder, campaign, product, etc.)
  • Visibility Class

By default, search results are sorted by relevance in groups of ten. Users can ask the AI application to return subsequent sets of results. If the initial request implies a need for recent assets (“Show me the latest approved campaign photos”), then results are returned by the latest modified date instead of relevance.

🎓

Insight: Objects

An object is a record type, such as an image, video, folder, user, or project. When you configure tabs, extensions, or metadata fields, you’re editing the object, and your changes then apply to every individual record of that type. For example, if you add a metadata field to a Brand object, the field appears on every brand record. Objects can have relationships. For example, a retail company might create relationships between a Product object and asset or project objects, so that each product record can be associated with multiple images, videos, catalogs, or marketing campaigns. Objects were previously described as asset types and subtypes.

Search results

Each returned asset includes a signed CDN link to the default preview format and a link to view the search results in the Orange Logic interface. Search results can also provide the following applicable metadata when requested:

  • Title
  • Unique identifier
  • File extension
  • Dimensions in pixels (if applicable)
  • File size
  • Creation date
  • Last modified date
  • Visibility Class
🎓

Insight: Signed CDN links

A content delivery network (CDN) is a set of globally distributed servers that store and deliver files, so content loads quickly for users anywhere. Orange Logic uses a CDN to serve asset files such as previews and downloads. The link you receive for one of these files is a signed CDN link.

"Signed" means the link includes a signature that grants access to one specific file. The signature reflects your Orange Logic permissions and any watermark applied to the asset, so the link shows only what you can already see. Access is built into the link, so it opens the file without a separate login and stops working once the signature expires.

MCP audit trail

Every record in Orange Logic has an audit trail that lists changes made to the record along with the user and timestamp. When someone uses an MCP connection to update a record’s metadata, role participants, or relationships, the change is logged in the audit trail as the AI application impersonating the user. For example, suppose Jennifer uses Claude to update a project’s status. The audit trail will list the change as completed by “Claude via impersonation.” The details note the Changed by user as Claude and includes an On behalf of line to represent the user.

A change in the audit trail performed by Claude on behalf of a user

Connect to Orange Logic’s MCP server

User authorization

An AI application does not access content in your Orange Logic environment until a user logs in and approves the connection. After that, every request the AI application sends runs as that user with that user's privileges.

Login and consent

Orange Logic's MCP server authorizes AI applications with OAuth 2.0 and the Authorization Code flow with Proof Key for Code Exchange (PKCE). PKCE adds a one-time check, so another application can't use an authorization code that it intercepts.

Before users log in, the AI application cannot access assets, metadata, or other content from your Orange Logic environment. By default, the AI application can establish an initial connection and discover basic server information before requesting sign-in. An Orange Logic administrator can submit a support request to require authentication when the application first connects. Include this ID in your support request: 2P7YOHELCAFJG.

A new connection follows this sequence:

  1. A user attempts to connect an AI application to your Orange Logic environment with an MCP connection.
  2. The AI application opens your environment's login page in the user's browser.
  3. The user logs in the same way they log in to Orange Logic. If your environment uses single sign-on (SSO), the login goes through your identity provider, and its login policies apply.
  4. The user approves the connection on Orange Logic's consent page.
  5. Orange Logic issues an access token that belongs to that user. The AI application sends the token with every subsequent request.

The AI application stores the authorization, so the user doesn't log in again at the start of each conversation. To end an AI application's access, the user or an administrator revokes the connection in Orange Logic.

🎓

Insight: OAuth and PKCE

OAuth is an open standard that lets a user give one application limited access to their account in another system without sharing their password. The user logs in directly with the system that holds the account and approves the request. That system then issues an access token to the application. The application presents the token with each request. Because the token belongs to one user, the application can do only what that user is allowed to do. When the user or an administrator revokes the token, the application loses that access.

Proof Key for Code Exchange (PKCE) is an extension to OAuth that protects the step in which an application exchanges an authorization code for an access token. Before the login starts, the application creates a random secret, called a code verifier, and stores it locally. It sends only a code challenge, which is a one-way hash of that secret, with the login request. After the user approves, the system returns an authorization code to the application. To exchange that code for an access token, the application must present the original code verifier. The system accepts the exchange only when the verifier matches the code challenge. An application that intercepts the authorization code doesn't have the code verifier, so it can't use the code. Current OAuth security guidance recommends PKCE for every type of application, including desktop and mobile applications that can't store a secret safely.

Privileges

The MCP server uses Orange Logic's security model, so it applies the same Security Functions and permissions as the Orange Logic interface.

Users need the SearchAPI Security Function to run a search using Orange Logic’s MCP server.

We do not recommend creating a shared account to authenticate MCP because every user would have access to the API account’s privileges, which could escalate an individual user’s access.

Set up MCP connections for your organization

If you have a team or enterprise plan, your account owner must authorize Orange Logic's MCP for your entire organization with your organization's URL: https://

If you have an individual Claude subscription, follow the process below to authorize MCP in the Claude desktop application and the claude.ai web application.

  1. Go to Settings > Connectors > Add custom connector.
  2. Enter https://<your-environment>.orangelogic.com/mcp and select Add. A browser window opens.
  3. Log in to Orange Logic and approve the connection on the consent page.

Manage MCP sessions

After you log in to Orange Logic and approve an MCP application’s connection, you can manage its authorization in your account settings.

Manage your sessions

Your authorized MCP applications are listed in your account settings in your user settings under Your authorized third-party applications.

To revoke an application:

  1. Go to user menu > My account.

  2. Click Your authorized third-party applications. The Third-party authorized applications window opens.

  3. Click the Revoke access button for the applications.

    Your authorized third-party applications window

    Your authorized third-party applications

Administrators: Manage user sessions

Administrators with access to a user’s account can revoke a user’s MCP session in the Orange Logic interface.

  1. Open the user’s account in System Settings.

  2. Select Your authorized third-party apps in the action menu.The Third-party authorized applications window opens.

  3. Click the Revoke access button for the applications.

    Open a user's authorized third-party applications in the actions menu

Credit consumption and cost

MCP is included with your Orange Logic package. MCP searches do not consume your AI credits, and there are no per-call charges. AI credits are consumed only if you trigger an AI service, such as an AI agent or AI media processing.

Deactivate MCP

If your organization doesn't want an active MCP endpoint, an Orange Logic administrator can submit a support request to deactivate MCP. Include this ID in your support request: 2P7YOHELCAFJG.

When MCP is deactivated, every request to /mcp returns 404 (Not Found). That response covers every host name your environment uses, including upload host names. It also covers variations of the path, such as /MCP and /mcp/.

Full MCP tool list

Foundation

  • Ping: Liveness check; the only tool visible before login.
  • WhoAmI: Returns the user the session is acting as (login, email, name, contact type).

Read

  • GetRecord: Resolves any identifier and returns a compact summary if readable. Unreadable or unknown IDs return as not_found.
  • GetUpdatableFields: Lists fields the user has permission to edit.
  • GetPossibleTaskStatuses: Returns the current status and available statuses. Requires an ID to search for the record.
  • GetAuthorityListValues: Lists controlled-vocabulary values for a dropdown field. The server-side filter is case- and accent-insensitive.
    1. Optional parameter: maxValues. The default is 25 values with a maximum of 100 values.

Write

  • UpdateFields: Batch edit fields per-record SUCCESS/ERROR/SKIPPED, one result per input in input order
    1. Parameters:
      1. taskIds[] (required, 1–20)
      2. assignments[] (required object)
        1. Field
        2. Mode: Overwrite | Prepend | Append | , ReplacedString (Replace only)
      3. augmentAuthorityLists (optional, default false)
  • SetTaskStatus: Sets a record status and returns the previous and new status. Requires an ID to search for the record.
    1. Required package: Creative Collaboration
  • PostComment: Adds a comment to the Comments & attachments extension on a record. Requires an ID to search for the record.
    1. Optional parameter: scope (private or public). The default is private.
  • SetParticipant: Adds, removes, or replaces a participant in a role. Requires an ID to search for the record.
    1. Required parameters:
      1. participant: Email address
      2. mode: Add, remove, or replace
    2. Required package: Creative Collaboration
  • RelateTasks: Adds or removes a relationship between two records.
    1. Required parameters:
      1. parentTaskId
      2. childTaskId
      3. relation: Relationship ID
      4. mode: add or remove