I am so excited to share an update with you on the development of a new MCP server for Tape. I explain in the video exactly what an MCP server is, what it’s meant to achieve, and a full demonstration of the creation of a brand new workspace and Tape system using the server.
I am eager to hear your feedback and thoughts. I want to give a HUGE shoutout to our hero @Jason for helping me test this. I have some folks who are testing it for me, and I’m hoping I can open it up to more testers once it’s a little bit more dialed in.
I have been privileged to have tried this over the last few days and I can quite simply say it is incredible. I have used it in Claude Desktop, VS Code and Zed using models other than Claude in the last two. So much is impressive about this tool but one of the things that I always find tricky to do is create and update calculation fields via the API this does it and I know Andrew has done a lot of work to tune this aspect and get it working smoothly.
Andrew says:
Well justifiably so - honestly this is one of the most useful and impressive uses of AI I have seen in a long time and possibly ever.
WOW
This is super incredibly awesome and impressive. Amazing job.
Thanks for sharing.
Do you anticipate turning this into an external tool that can be purchased and bundled with Tape for users in different Orgs?
Is it set up in a way that users can tap into it separately and still maintain end goal focus for that individual users use case without being clouded by goals stated by other users? Or will each user have to set up an independent version of the MCP server to execute their specific wishes?
As for the rest of your questions, based on how the Tape API works, what I have set up is basically one API token per user account in my system. You will get a code that will correspond to one API token, and everything will be contained within that one tenant ID. So whatever you can get done with one API token will be the answer to your question, I hope.
To answer some of your questions in a more specific fashion, you actually ask some really smart questions and these are some of the things I’m thinking thinking about.
For example, if you were a tenant in my system, imagine using the MCP server to index the entire system app by app and providing a visual interface similar to QuivvyTools where user could see information about specific fields, and make decisions about what updates they want to make. It’s interesting you use the phrase “end goal focus”… imagine a Github repo that is a collection of JSON files for each app in the system, a team uses this Github repo to branch out their development work and use the MCP server to make the updates, updating the JSON representation of each app along the way and creating pull requests for their work. Essentially equivalent to how professional development teams work today.
If Github was not your preferred method of sharing code, imagine setting up a Tape system where a user authenticated with Claude Code, made some significant changes and then ended their session, and then there was a custom tool in the MCP environment that would summarize what they worked on in that session and create a Tape/Podio/JIRA/whatever item somewhere relevant to document the experience.
This MCP server in a controlled environment with your model of choice makes all of this possible.
@andrew.cranston I love that idea.
This ‘dev session summary’ example would actually be far more useful to me than the example of building out brand new applications within Tape.
By having the MCP server maintain a plain english change log or visualization of how a system evolves over time, this could help so so much with troubleshooting errors and accountability if multiple admins have the ability to make major edits.
Tape doesn’t currently have an endpoint for tracking app revisions, as an example. However, my MCP server is tracking every single tool call that every single user is making. So technically, I could see an entire tool history for a particular user if I wanted to.
