Gen AI is taking over the developer world. From prompt engineering and code assistants to Vibe Coding, it has fundamentally changed how developers build and think about Software. Quickly becoming a must-have tool in every developer’s toolkit.

Gen AI has enabled incredible speed and is making the art of building software accessible to more people. And with the explosion of AI tools comes the question of unifying approaches.

A standard approach makes it easier for builders to use different tools and more manageable for the makers of the AI tools to stay adaptable and extendable.

Model Context Protocol (MCP) as Cutting-Edge Tech

Enter the Model Context Protocol (MCP). This protocol describes a standard way for AI models to access relevant information beyond the local context.

Essentially, MCP allows large language models (LLMs) to use external data beyond their training data and the immediate local context, making them much more useful as coding assistants.

For example, imagine debugging a persistent error with Cody AI, and your codebase uses an external service. Normally, Cody would only have access to your code context and its training data to help you.

But with MCP, Cody would be able to read the service’s docs and even search for relevant issues in the service’s public forums, using that as extra context, immediately making it a more capable assistant.

The Role of Supabase

Supabase has been incredibly quick to adapt to the changing tides and positioned itself as the perfect backend solution for AI applications.

They championed a Postgres native solution, PgVector, as the ideal Vector store at a time when Vector DB providers were seen as the go-to. They managed to shift this and have developers just use Postgres<sup>TM</sup> 🤭 for their vector needs.

Now, they are at the forefront of the low-code AI movement, embracing tools like Bolt.new and Lovable.dev. They have done this by creating seamless integrations and pushing high-quality educational content.

Why Supabase

I chose Supabase for this exercise because it has done an amazing job of staying on top of tech trends despite being based on old, stable, and maybe even boring tech (Postgres).

It’s commendable that despite having deep roots, Supabase also has branches and leaves that stretch to the sky.

Plus, I’m part of their community and already quite familiar with the product.

How I’d Approach Enhancing Supabase’s DX Around MCPs

As a developer experience engineer at Supabase, I would focus on these areas in light of MCPs as emerging tech.

Deepening Documentation

I would focus on improving the MCP documentation page and ensuring that Supabase mentions its own MCP server, which can interact more with Supabase projects than the listed read-only Postgres one.

I would also maintain a list of known issues or link to the issues tab of the MCP public repository.

Finally, I would link official videos explaining MCPs to the docs page so it’s clearer for developers.

Tutorials and Content

I would create technical tutorials on MCPs, including definitions and how-tos, as they are still new.

I would collaborate with the rest of the DevRel team to cover all main environments, tools, and use cases that would benefit from an MCP setup.

The Proof in the Pudding

And because I know that Supabase culture prioritizes actions over yapping, here is me starting the work I listed above:

AI, MCPs, and How I'd Approach DX at Supabase

Table of contents