Hello users, I have begun work on an MCP server for Survey Solutions Headquarters and am reaching out here for any interested users who would like to collaborate on the development / testing of this project. Here’s some details below about what I am planning:
What is an MCP server? - An MCP server is a layer which sits between an applications API access and an Agentic AI system (such as Claude co-work, Chat-GPT Codex, Openwork, etc). It essentially simplifies the raw API calls into tools and workflows the AI Agent can call more easily understand and call without having to develop full programs to call each API function.
Why make an MCP for Survey Solutions? Two big reasons
-
Security - At present the API access for Survey Solutions does not have different levels of access - If I grant an AI agent straight API access it can make any change or deletion to the server. The MCP server will have changeable levels of access for the AI Agent - read only, fieldwork qc, full write and delete.
-
Functionality - Rather than relying on the AI agent to understand the full Survey Solutions ecosystem and then work out which API calls would work to achieve a purpose, the MCP wraps up the API calls into more self explanatory tools the AI agent can leverage. Additionally it can include common workflows which combine several API calls together: For example a workflow which starts the export, uses polling to check the status and downloads when done.
The goal is to get a secure, safe connection between your AI Agent software and a Headquarters instance, where the LLM is well equipped with tools to handle most common tasks in Headquarters. For example when I need to share a test link of a questionnaire with a non-technical user who doesn’t have a designer account I will make a testing workspace, import the questionnaire, create dummy interviewer and supervisor accounts, create an assignment with a -1 web interview, download the web links and finally have a test link i can email to somebody. I’m aiming to have the MCP connection make this workflow as simple as telling the agent - “make me a new test link for x questionnaire” So I am interested in hearing from users laborious little workflows like this which would be cool to automate or if you are interested to collaborate in the design / testing please drop me a message.
Thanks!
Lachlan