MCP

MESA MCP lets you supercharge your AI prompts by calling MESA workflows to fetch information and even take action.

Get started by creating a workflow in MESA with an MPC trigger. Then, let your AI app know about MESA MCP and start chatting. In the steps below, we use Claude Desktop, but MESA MCP also supports Cursor and many other AI apps.

Creating a MESA MCP workflow

1. Create a new workflow in MESA. Search for and select MCP to add a MCP trigger.

2. Configure the MCP step with any "parameters" your MCP workflow will need. If you want to search Shopify Orders, for instance, you might want to add parameters like status, created_at_min, created_at_max that will match the types of queries you will want to make. Make sure to give these parameters useful, explicit descriptions for the AI tools to interpret.

3. Add any additional steps you would like your workflow to perform. Make sure to use any parameters you created to outfit your step(s) with the information they need to perform actions for your AI tool.

4. Name the workflow something AI tools can understand, eg, "List Shopify Orders with Query". Be as explicit as possible to help AI tools determine which of your workflows to use. For instance, if you also wanted to provide your AI tool with Shopify orders by a single customer, you could create a new workflow "List Shopify Orders by Customer Email".

5. Give the workflow as descriptive of a description as possible. If you have similar "sounding" workflows, the description can provide context to the AI tool about how to decide which workflows(s) to choose.

6. Enable the workflow.

7. Copy the connection information provided by the MCP trigger step.

   Note: you only need to configure Claude or other AI tools to connect to your MESA MCP server once.

Using MESA MCP with Claude

Get Claude Desktop

Currently, Claude can only talk to MESA's MCP workflows with Claude Desktop, so make sure that application is installed on your computer. This article on MCP for Claude Desktop has other potentially useful information.

Ensure node is installed on your computer

Next, you will need to make sure that Node is installed on your computer. To check if it is installed, open your "terminal" program on your computer and run this command:

node --version

1. If you see command not found: node it means you will have to install Node.
2. If you see any version less than 20.1.0 it means you will need to update your system's node version to at least 20.1.0.

Connect Claude to your MESA MCP server

1. Click on “Developer” in the left-hand bar of the Settings pane, and then click on “Edit Config”.

2. This will create a configuration file at:

   • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json

   • Windows: %APPDATA%\Claude\claude_desktop_config.json

3. Edit the claude_desktop_config.json file with the connection info provided by the Skill trigger. It will look something like this, though your information will be inserted where ID and KEY are below:

   {
     "mcpServers": {
       "mesa": {
         "command": "npx",
         "args": [
           "mcp-remote",
           "https://mcp-server.getmesa.com/sse/ID/KEY"
         ],
         "env": {}
       }
     }
   }

4. Save the file and restart Claude Desktop.

   Note: you only need to configure Claude or other AI tools to connect to your MESA MCP server once.

Call MESA MCP workflows in Claude

Upon restarting, you should see a hammer icon in or near the input box. Clicking that hammer will show your Available MESA MCP workflows!

Not seeing the hammer icon? Take a look at Claude's MCP troubleshooting section.

Try asking Claude something that would require information provided by your workflow eg. "Find Shopify Orders from March 2025". If Claude decides that it needs to run your workflow, you will be prompted for your approval to use the tool.

Technical Notes

• Treat your MCP Configuration JSON like a password. It can be used to access and update your data.

• The maximum request time is limited based on the Task compute limit of your plan. There is also a limit of 60 seconds per request, even if your plan allows longer requests. Requests that take more time to execute will be timed out, and no response will be returned.
• Requests are rate-limited based on the Incoming rate limit of your plan.

Common Issues

Getting Claude Desktop running can be difficult, and the error messages can be unhelpful. Here are some solutions to common issues we have encountered.

Claude's "response was interrupted"

Usually, this means that too much information in the chat has already been processed, and Claude is running into issues with the "context window" size.

Some strategies:

1. Start a new chat. Sometimes this can help, but if you're requesting a lot of data on every run, it may not do the trick.

2. Upgrade your plan. Claude Pro includes a much larger context window than the free version.

3. Reduce the size of your workflow's response. Depending on the app you may be able to configure your steps to return less information. This could be in the form of "only return these fields" or simply "limit number of responses". You could also use Loops or Custom Code to reduce the size of the information you respond with.

Claude still can't access my MESA MCP workflows

As noted above, you need a version of Node newer than 20.1.0 to access MESA's Skills.

Next a look at Claude's MCP troubleshooting section for anything you might have missed or misconfigured.

If you are a more technical user and have a node "version utility" like nvm installed. It's possible that Claude is getting confused with node versions as npx runs.

The mcp-server-mesa.log file may have things in it like:

npx: installed 120 in 8.27s
Unexpected token {
}

If that is the case, you may need to provide Claude with more information to function properly. These steps are only valid for Mac users but something similar should be possible for Windows:

1. Open your terminal application and run:

   which npx

2. Replace npx in the "command": "npx", of the Claude configuration file with the full path, eg. "command": "/Users/full/path/node/22.14.0/bin/npx",.

3. In your terminal application run:

   echo $PATH

4. Add the PATH "env" variable to the Claude configuration file for your MESA MCP server, and paste in the response from echo $PATH eg.

   {
     "mcpServers": {
       "mesa": {
         "command": "/Users/full/path/node/22.14.0/bin/npx",
         "args": [
           "mcp-remote",
           "https://mcp-server.getmesa.com/sse/ID/KEY"
         ],
         "env": {
           "PATH": "/The/Path/Returned/By/Your/Terminal"
         }
       }
     }
   }

5. Restart Claude Desktop.

I saw my Skills, but they went away after I restarted Claude Desktop

You might need to kill the mcp-remote process running on your computer. In your terminal application run:

pkill -f mcp-remote