2026-06-01

by Zachary Siciliani

I recently came back from TDX, Salesforce’s developer conference, and shared a recap with our team. Like most conferences, there was a mix of genuinely useful signals, early ideas that are not quite ready, and a fair number of buzzwords.

Rather than listing features or announcements, I want to capture what stood out from a developer’s point of view and what direction Salesforce is taking its platform.

Salesforce as a backend platform

The most consistent theme across the keynote and sessions was what Salesforce calls “Headless 360”. The core idea is that Salesforce does not need to be the primary interface where work happens.

More and more capabilities are being exposed through APIs and MCPs, with the assumption that users will interact through Slack, Teams, or AI-driven tools. Salesforce becomes the system of record and execution layer, not the destination itself. Anything you can do in the Salesforce UI should also be possible externally without needing to go through it first.

Is it fully there yet? No. Reports were barely addressed, and many examples were simplified. But the direction is clear, and it is something developers should pay attention to.

Giving AI real context

A lot of attention was given to Model Context Protocols. MCPs allow external tools to access Salesforce metadata and understand the structure of an org.

From a development perspective, this is significant. AI becomes much more useful when it knows object models, field names, and relationships. Generating code or changes without that context is limiting. With it, the output is closer to something usable.

In theory, you can connect tools like Cursor to Salesforce through an MCP and let the AI build objects, flows, or code with full awareness of the org. There are still open questions, especially with consultants working across multiple orgs, but the concept itself is solid and worth experimenting with.

AI-assisted development today

AI is clearly effective when working with code, particularly modern JavaScript frameworks such as React. It accelerates development, helps navigate unfamiliar codebases, and reduces friction for repetitive tasks.

Flows are a different story. AI can read and assist with existing Flows, but building complex, deployable Flows from scratch is still unreliable. For now, AI works best as an assistant rather than an author in this area.

Understanding these limits is important for setting expectations and deciding where AI adds value today.

React inside Salesforce

Support for building React applications directly in Salesforce is one of the more tangible changes.

This enables fully custom interfaces that are not constrained by the Lightning look and feel. It is especially relevant for portals, branded experiences, and UX-heavy applications. For standard internal workflows, existing tools remain sufficient.

It also aligns well with AI-assisted development, since React is an ecosystem where AI performs particularly well.

DevOps and delivery

The sessions on DevOps Center were less exciting on the surface but more relevant to my day to day as a developer. Faster deployments, better integration, and more API access all make my daily work smoother.

Salesforce also highlighted testing integrations with external tools. These are powerful, but often priced for organizations with frequent, high-volume deployments. For many teams, solid Git practices and thoughtful release management still provide most of the value.

Maintainability still matters

One recurring discussion was whether AI changes the balance between Flows and code. In practice, maintainability remains the deciding factor, despite AI considerably lowering the barrier of entry to writing code.

When solutions are handed over to clients, clarity and long-term support matter more than how quickly something was built. Flows are often easier to understand and maintain for non-developers, and AI does not change that reality.

Closing thoughts

TDX reinforced Salesforce’s direction more than it introduced finished solutions. The platform is moving toward greater flexibility, AI-assisted development, and less reliance on its own UI.

For me, the value was in seeing where Salesforce is investing and how those choices will affect developers over the next few years. It helped sharpen priorities and identify which areas are worth exploring now, and which are better watched from a distance.

I would go back. While the hype was nice, the clarity and direction are always worthwhile to witness first-hand.