> For the complete documentation index, see [llms.txt](https://developers.make.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://developers.make.com/mcp-server/make-mcp-server.md). # Make MCP Server ## Make MCP server Make MCP server allows AI systems, such as large language models (LLMs), to run scenarios and manage your Make account. Model Context Protocol (MCP) is a communication standard between AI and external systems that enables safe interactions by defining endpoints and providing authentication. ### How Make MCP server works AI systems like Claude and ChatGPT act as MCP clients of Make MCP server. The server gives them access to MCP tools that: * Run **active** and **on-demand** scenarios. * View and modify scenarios and their related entities, such as connections, webhooks, and data stores. * View and modify teams and organizations. Scopes determine the [tools available](#view-available-tools). Scenario run tools are available to all plans, and management tools are available to paid plans. You select scopes when you [connect Make MCP server to an MCP client](#connection-methods). ### Make Skills Make Skills are Markdown files with instructions on building scenarios, configuring modules, and connecting to Make MCP server or third-party services. You can install them in MCP clients to connect to Make MCP server without setting up the connection manually and help clients reliably perform Make-specific tasks. For installation instructions, see [Make Skills](https://skills.make.com/). ### Connection methods Make MCP server is a Make-hosted, cloud-based server that runs via stateless Streamable HTTP. Connect it to MCP clients using OAuth or an MCP token. Alternatively, install Make Skills, which include connection details for Make MCP server. #### OAuth Use the following connection URL: ``` https://mcp.make.com ``` For more information, see [Connect using OAuth](https://developers.make.com/mcp-server/connect-using-oauth). #### MCP token The URL you use depends on whether your client supports adding the MCP token in Authorization headers. **MCP token in URL** Use the following connection URL, and replace `` and `` with actual values: ``` https:///mcp/u/ ``` * `MAKE_ZONE`: The zone your organization is hosted in. For example, `eu2.make.com`. * `MCP_TOKEN`: Your MCP token. To obtain one, see [Connect using an MCP token](https://developers.make.com/mcp-server/connect-using-mcp-token). Optionally, append the following query parameter to the URL to set the tool name length: * `?maxToolNameLength=`: Generated tool names have a maximum length of 56 characters by default, to be compatible with most AI systems. Use this parameter to set a different maximum length. The allowed range is 32 to 160 characters. {% hint style="info" %} Optionally, to control which scenarios are available as tools through your MCP token, see Scenarios as tools access control. {% endhint %} **MCP token in header** If you add the MCP token in an Authorization header, use the following URL: ``` https:///mcp ``` Specify the MCP token in the Authorization header: ``` Authorization: Bearer ``` ### MCP tools MCP tools are functions on Make MCP server that MCP clients call to interact with Make, such as running scenarios or updating data stores. #### View available tools You can view the tools on Make MCP server in a few ways: | Method | Action | Shown | | ---------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------ | | MCP client settings | In your MCP client's connector settings, select Make MCP server. | Tool names, and in some clients, descriptions and categories | | MCP client command | In your MCP client, run its MCP command, for example `/mcp` in Claude Code, then select Make MCP server and a tool. | Tool names, descriptions, and schemas | | MCP Inspector | In a terminal, run `npx @modelcontextprotocol/inspector --cli https:///mcp --transport http --header "Authorization: Bearer "`. | Tool names, descriptions, and schemas | | `tools/list` request | In your app or agent's code, send a `tools/list` request to Make MCP server, or use an SDK method, such as `list_tools()` (Python) or `listTools()` (TypeScript). | Tool names, descriptions, and schemas | | Prompt in AI assistant | In your MCP client's chat, ask something like "What tools do you have from Make MCP server?" | Tool names and descriptions | #### Tool call timeout limits Tool calls from MCP clients to Make MCP server have a timeout for returning output. Limits depend on the tool type and connection method. **Scenario run tools** * `https://mcp.make.com` (OAuth): 25 seconds * `https:///mcp/u/` (MCP token): 40 seconds **Management tools** * `https://mcp.make.com` (OAuth): 30 seconds * `https:///mcp/u/` (MCP token): 60 seconds **Retrieve outputs after timeout** If a scenario run tool call times out, the scenario continues running in Make for up to 40 minutes. The timeout response includes an `executionId` that an MCP client can use to get the output after the scenario finishes. Example response: ``` { instruction: 'The Tool execution has started, but did not complete yet. Use a corresponding Tool for getting Execution Results to check for the result asynchronously.', executionId: '4eb841b3fe3d48a5b2b59e2eed3f55be', scenarioId: 1, }, ``` Select the `scenarios:read` scope to allow an MCP client to retrieve outputs after a timeout once a scenario run finishes. #### Scenario inputs and outputs When an MCP client uses a scenario as a tool, it refers to the scenario's [inputs and outputs](https://help.make.com/scenario-inputs-and-outputs) to know what data to send and what it gets back: * **Inputs** are parameters that an MCP client fills with data when the scenario runs. * **Outputs** are data the scenario returns to the MCP client. {% hint style="info" %} Write detailed scenario descriptions to help MCP clients understand what each scenario does and get more reliable results. {% endhint %}