Missing a connector? Build it — or have AI draft it

A connector catalog is only as good as its long tail. Sooner or later you hit the internal API, the niche SaaS, or the bespoke protocol that no packaged node covers — and on most platforms that is where automation stops and a support ticket begins. FluidGrids treats the missing connector as something you build, not something you wait for.

You have two ways to start. Write the node package yourself when you want full control, or describe the connector in plain English and let AI draft it — name the language, the node category, and a slug, and it produces a first-pass package for you to review. Either path ends the same way: a candidate node ready to enter the workspace's catalog rather than a snippet of code with nowhere to live.

Getting there is a governed lifecycle, not a raw upload. Registering your package brings it into the system as a submission, and the submission moves through validation and review before it can be trusted — you can watch each submission's state, retry one that failed, and clear a dead attempt. Nothing reaches the palette on a hope; it reaches the palette because it passed.

Once approved, installing the connector drops it into the workspace catalog, and an admin can promote, deprecate, or retire it over its lifetime — and uninstall it to pull it back out. From that moment your custom node behaves exactly like a built-in one: it appears in the palette search, renders its own Adaptive-Card configuration form from its manifest, references workspace connections for its secrets, and pins its version so existing workflows keep running the shape they were built against.

The payoff is that the catalog stops being a fixed menu and becomes a surface your team extends. The connector that used to block a whole automation becomes a package you author once, review once, and install once — and every workflow in the workspace can then drag it onto the canvas like it was there all along.