The painful part of most integration work isn't the logic — it's the archaeology. Which endpoint? Which fields are required? What format does the timestamp want? FluidGrids moves that knowledge into the node itself, so configuring an integration feels like filling a card, not reading API docs.
The builder ships 30 manifest-driven store nodes — Slack, GitHub, Stripe, Google Workspace, OpenAI, Anthropic, Salesforce, and more — alongside 21 built-in flow-control nodes, with further connector packages being vendored in. Search the palette by name, category, or keyword; each node carries its icon, description, and operations straight from its manifest.
Click a node on the canvas and pick an operation — "send message", "create row", "post data". The configuration panel renders an Adaptive-Card form generated from the manifest's declared input fields: required flags enforced, defaults pre-filled, types respected. No per-connector frontend was hand-built, which means every connector's config experience is consistent and every new connector arrives fully formed.
Nodes install per workspace, so admins control which connectors a workspace can use. Versioning is first-class: nodes are versioned packages, workflows pin the versions they were built with, and a deprecated node version gets flagged in the builder while historical workflows that reference it still render.
When an operation needs credentials, the card surfaces a connection picker filtered to matching providers — wired into the connections system, so secrets never appear in the form. Fill the card, save, and the configuration persists inside the workflow definition, surviving every reload.
Do it yourself
Configure any connector in the palette by searching for it, dropping the node, and filling the Adaptive-Card form generated from its own manifest — no API docs required.
Open a workflow in the builder so the node palette is visible.
You should see: The canvas and the searchable node palette load.
Ready to make this your story?
