Model Context Protocol
MCP toolboxes
10 min
mcp toolboxes are dedicated {{product name}} mcp servers that allow mcp clients, such as claude and chatgpt, to access and use your {{scenario plural lowercase}} they consist of a specific set of mcp tools (scenarios that you've set to active with on demand scheduling) in this guide, learn when to use mcp toolboxes, their key features, and how to work with them in {{product name}} and mcp clients when to use mcp toolboxes use mcp toolboxes when you want more control over the tools available to mcp clients you want a secure way to share mcp tools with team members and clients make mcp server is another way to use mcp with make to learn more, see docid 95jjtreuq1qmryibxvbkx key features mcp toolboxes offer these features tool management in {{product name}} (add, configure, and delete) token based authorization using keys a unique url per toolbox the option to create multiple keys per toolbox for sharing configurable tool labels, descriptions, and types ( read only or read write ) create a mcp toolbox to create an mcp toolbox, including its unique url and key in {{product name}} , click mcp toolboxes on the left sidebar, then create toolbox at the top name your toolbox select the tools you want to have in your toolbox only active {{scenario plural lowercase}} with on demand scheduling appear in the tool list click save in the create key dialog, copy the key provided and store it in a safe place click close in mcp server url , copy the url provided you've now created an mcp toolbox in {{product name}} connect your mcp toolbox to an mcp client before connecting your mcp toolbox to an mcp client, you must choose an mcp client typical options include claude, chatgpt, and cursor check what the client supports in the client's mcp server documentation, check if the client supports authorization headers and the transport methods available http streamable is the recommended method due to its connection reliability next, continue to step 1 to configure your url based on what the client supports step 1 configure your url you must have a complete mcp server url to connect to an mcp client to configure the url configure your url based on the following structure, depending on whether the mcp client provides an authorization header with authorization headers \<mcp toolbox url> with the key in the authorization header header name authorization header value bearer \<toolox key> without authorization headers \<mcp toolbox url> /t/ \<toolbox key> / \<transport method> replace the placeholders with your actual values mcp toolbox url the url generated when you created the mcp toolbox toolbox key the key generated for the mcp toolbox transport method the path of your preferred transport method stateless streamable http ( /stateless ) is recommended due to its connection reliability example with authorization headers url https //eu2 make com/mcp/server/••••••••••••• https //eu2 make com/mcp/server/•••••••••••••/t/•••••••••••••/stateless ( mcp toolbox url ) authorization header ••••••••••••• ( toolbox key ) example without authorization headers url https //eu2 make com/mcp/server/•••••••••••••/t/•••••••••••••/stateless ( mcp toolbox url /t/ toolbox key / transport method ) you've now configured your mcp toolbox url you'll add it when you're connecting to an mcp client step 2 connect to any mcp client to connect a mcp toolbox to an mcp client in your mcp toolbox, in mcp server url , click connect with and select an mcp client if you don't see your preferred client, open its documentation for adding remote mcp servers in the documentation, begin following the steps for connecting to your client when asked for a url, add the url that you configured in step 1 if applicable, add the key in the authorization header finish the steps you've now connected your mcp toolbox to mcp client manage your mcp toolbox once you create an mcp toolbox, you can add or edit tools, add keys, and delete the toolbox add tools to add tools to an existing mcp toolbox in mcp toolboxes , click an existing toolbox in tools , click add search or select a tool or multiple tools click save you've now added tools to an existing mcp toolbox configure tools optionally, configure your tool's label, description, and behavior label and description give your tool a custom tool name and description when its existing ones in {{product name}} may confuse an mcp client or prevent it from choosing the tool correctly by default, the client sees existing names and descriptions, but custom ones replace them behavior select read only or read write as tool annotations in mcp clients that support them, annotations help clients understand how cautious to be with a tool, such as asking for permission before running it by default, tools are read write mark a tool as read only if it only reads data rather than making changes, such as adding or deleting this option helps the client run your tools more quickly to configure tools in mcp toolboxes , click an existing toolbox in tools , click the settings icon in label , enter a custom tool name, for example, "watches unread emails " alternatively, leave the field empty to use the default name in description , enter a custom tool description, for example, "watches inbox for unread emails with a lead label " alternatively, leave the field empty to use the default description in behavior , select read only or read write from the dropdown click save you can go back to configure your tools at any time add keys each mcp toolbox can have multiple keys to share or use across mcp clients to add keys to your mcp toolbox in mcp toolboxes , click an existing toolbox in keys , click add in name , name your key so you can identify it save the key in a safe place so you can share it later click close you've now added a key to your toolbox to delete a key, click the x symbol next to the key to delete delete an mcp toolbox to delete an mcp toolbox in mcp toolboxes , click the three dot menu next to an existing toolbox click delete click delete to confirm you've now deleted an mcp toolbox mcp toolbox timeout your mcp toolbox scenarios must finish running within 40 seconds if a scenario runs longer, the connection times out and no outputs are returned to the client