Query, write, and call Postgres functions in a Supabase project over its auto-generated PostgREST REST API.
Supabase ships in the w6w first-party pack. It declares 7 actions, 2 health checks, and the host runs its code in a sandbox that never sees the credential.
io.w6w.supabaseSupabase is a hosted Postgres platform, and this app talks to a project’s data over its REST API rather than the Postgres wire protocol — list rows from a table with PostgREST’s own filter syntax, fetch a single row, count matching rows without transferring them, insert or upsert rows, update or delete matching rows, and call a Postgres function exposed through the API.
Every action is table-agnostic, since a project’s own schema isn’t known ahead of time — a workflow names the table (or function) and supplies the filter and payload it needs. Write actions return the affected rows by default, so there’s rarely a need for a separate read afterward to see what changed.
Which key to use matters: the anon key respects whatever Row Level Security policies a project defines, while the service-role key bypasses them entirely and should only sit on a connection that genuinely needs full read and write access, never one exposed to untrusted input like a public form or webhook. A fit for automating reads and writes against a Supabase-backed app’s data, or calling a stored Postgres function as part of a workflow.
Three routes to the same 7 actions. The Workflow tab is generated from Supabase's own manifest and carries its real ids, so it is copy-pasteable; the Code and CLI examples are the same call for any action on any app, so every app-specific value in them is a blank you fill in.
rows-count Count matching rows without transferring them, via `HEAD` + `Prefer: count=exact`.
rows-delete Delete every row matching a filter. `filters` is required so a blank form can't accidentally empty the whole table.
rows-list Query rows from a table or view, with PostgREST filter/select/order/paging syntax.
rows-update Set columns on every row matching a filter. `filters` is required so a blank form can't accidentally rewrite the whole table.
rpc-call Call a Postgres function exposed via PostgREST's `/rpc/<function>` endpoint.
A workflow step names the app and the action, and the editor fills in the
connection when you pick one. This is the Step shape from the
workflow spec, carrying Supabase's real ids.
{
"manifestVersion": "2",
"name": "supabase-example",
"steps": [
{
"id": "row-get",
"uses": {
"app": "io.w6w.supabase",
"action": "row-get",
"connection": "conn_YOUR_CONNECTION_ID"
},
"with": {
"table": "<table>",
"filters": "<filters>"
}
}
]
}row-get rows-count rows-insert rows-list rows-update +2 more actions available
Every app-specific value here is a blank you have to fill in. An
app action is reached through the connection that authenticates it, so the
address is a connection id, not the app id — and connections belong to your account,
so a public page cannot know yours. Create one for Supabase, then fill in
the three blanks: conn_YOUR_CONNECTION_ID, the action key, and the
parameters that action declares. The call itself is real — the shape is transcribed
from the studio's own snippet builder, which prints the same kind of blanks — but
nothing in it is specific to Supabase. The Workflow tab is where this app's
real ids are.
npm install @w6w/sdkyarn add @w6w/sdkpnpm add @w6w/sdkdeno add npm:@w6w/sdkimport { W6wClient, isActionRun } from "@w6w/sdk";
// Reads W6W_BASE_URL and W6W_TOKEN from the environment when omitted.
const client = new W6wClient();
const envelope = await client.run({
urn: "conn_YOUR_CONNECTION_ID",
action: "rows-insert",
payload: {
table: "<value>",
rows: "<value>",
// upsert: "<value>",
// onConflict: "<value>",
// select: "<value>",
},
});
if (isActionRun(envelope)) console.log(envelope.value); npm install -g @w6w/cli w6w run conn_YOUR_CONNECTION_ID --action rows-insert --payload '{"table":"<value>","rows":"<value>"}' Give an AI agent Supabase — without giving it Supabase's credentials. One MCP endpoint exposes every app, function and workflow the caller is entitled to, as tools it can discover and run. Access is granted per team while we onboard.
One tool call{
"name": "w6w_invoke",
"arguments": {
"ref": "app:io.w6w.supabase#row-get",
"input": {
"table": "<table>",
"filters": "<filters>"
}
}
}
Every tool names its target with a single ref. The
app: form above doesn't name a connection at all — the
host resolves which of the caller's Supabase connections to sign
with, and refuses rather than guesses when the answer is ambiguous.
The token is attached host-side, at the moment of the call. It is never a tool argument, never in the model's context, and never in a transcript — so a prompt injection has nothing to exfiltrate.
Tools are derived per end user from what that person has actually connected and is entitled to — not one shared bot identity carrying the union of everyone's access.
Multi-step work runs on the workflow engine and returns a run handle the agent can poll — retries, branching and state survive the conversation that started them.
Supabase's declared health checks are on the surface too, so an agent can tell "the vendor is down" from "your credential expired" before it burns a retry on either.
The MCP surface is part of the hosted platform. Supabase itself is MIT, and the runtime that executes it is source-available (FSL).
Supabase declares its own checks, so its health is a property of the app rather than something the host guesses at.
Open incidents from Supabase's Statuspage.io status feed. State is read from each incident's leading status word (Investigating/Identified/Monitoring/Update/Resolved), not from the newest headline.
Unauthenticated request to this connection's Supabase project host. A 401 passes — it proves the gateway is serving; credential validity is the `auth:*` check's job.