What is MCP?
MCP (Model Context Protocol) is an open-source standard for connecting AI applications to external systems. Using MCP, AI applications like Claude or ChatGPT can connect to data sources, tools, and workflows, enabling them to access key information and perform tasks.
Enterprise MCP is the Workato component that allows you to create and manage MCP servers that give MCP clients access to your Workato automations, including project assets and API collections. Workato hosts the servers and enforces the permissions of the authenticated user.
Before you open a ticket
First, check status.workato.com for an ongoing incident. If there is one, follow its updates and check the affected functionality again once it is resolved.
The following actions often resolve the most common MCP issues, or help you work around them:
1. Check access and authentication
Open your MCP server in AI Hub > Enterprise MCP > Custom servers and check which access method it uses on the Settings tab.
- Token: Confirm that the MCP URL and token are correctly configured in your MCP client. Switching a server to Workato Identity revokes its token, so reconfigure your clients if the method was changed.
- Workato Identity: Access is not granted automatically. Confirm that the user belongs to an end-user group, and that the group is added to the server’s Settings > End user access page.
2. Check the server limits
If all requests fail, open Settings > Limits. A rate limit or usage quota may be used up, and then all requests are blocked until the interval passes or the quota resets. Also, check the Allowed IPs and Blocked IPs lists on the same page. Blocked addresses take precedence over allowed ones.
3. Retry the request and reconnect the client
AI responses can vary between attempts, even for the same request. Repeat the request a few times, including in a new chat. If it still fails, remove the MCP server from your client and add it again.
4. Test the server with MCP Inspector
MCP Inspector is an open-source tool from the MCP community that allows you to test and debug an MCP server without an AI client. It shows whether the issue is in the server or in the client.
- Download and install Node.js
- Run
npx @modelcontextprotocol/inspectorand open the link it prints in your browser - Set Transport Type to Streamable HTTP
- Enter the MCP URL of your server. For token-based access, include the token. For Workato Identity access, do not include a token, you will be prompted to log in via the Workato authorization page instead.
- Click Connect, list the tools, and run the tool that fails
If the tool works in MCP Inspector but not in your AI client, check the client configuration. If it fails in both, include the error message in the ticket. For more details, see the MCP Inspector documentation.
Information to include
Including the following information in your ticket helps Workato Product Support resolve it faster:
- Details of the issue
- Remote MCP URL
- Workspace ID
- Screenshots
- MCP client
How to collect it
Details of the issue
Describe the following:
- Detailed steps to reproduce the issue
- What you were trying to achieve and the current result
- The exact error message, if any
- Timestamp of the last occurrence, with time zone
- Any recent changes
- Whether it worked before or is a new setup
Remote MCP URL
The URL contains the ID of your MCP server. It helps Product Support identify the affected MCP server.
- Go to AI Hub > Enterprise MCP > Custom servers and click your server
- Switch to the Details tab
- Copy the Remote MCP URL value
URL example: https://XXXXX.apim.mcp.workato.com/
Workspace ID
Each environment (Development, Test, Production) has its own workspace ID. Provide the ID of the environment where the affected MCP server is configured. To find it:
- Open the environment where the issue can be reproduced
- Navigate to Workspace Admin > Settings > General
- Copy the Workspace ID value
Screenshots
In the Workato UI, go to the page where the issue occurs and take a screenshot of the full browser window, so that the error and the address bar are both visible. In the MCP client, take a screenshot of the chat where the issue occurs.
MCP client
MCP client with its name and version (Claude, ChatGPT, Cursor, and so on). It can help Product Support to reproduce the issue.
Comments
0 comments
Please sign in to leave a comment.