Skip to content

Stacksync × Supabase Miami: AI-Native OperationsOct 13Reserve your spot

Your connection path

PostgreSQL
Claude Code
Stacksync MCP OAuth

Your assistant works through your Stacksync workspace and its available connections.

First request

“Prepare a bounded operational query”

Stacksync MCP · Database

Connect PostgreSQL to Claude Code with MCP.

Connect an AI assistant to Stacksync to inspect a PostgreSQL connection, prepare bounded SQL queries and plan app-to-database syncs. Query access and sync management are separate operations, with different permission and configuration requirements.

Connector actionsSync management

01 / The connection

How PostgreSQL works with Claude Code

Connect an AI assistant to Stacksync to inspect a PostgreSQL connection, prepare bounded SQL queries and plan app-to-database syncs. Query access and sync management are separate operations, with different permission and configuration requirements. Add the Stacksync MCP server to Claude Code, then use the connection path below.

The server belongs to Stacksync. The available PostgreSQL operations depend on your connection, permissions, and connector modules.

Explore the PostgreSQL connector →

02 / Capabilities

PostgreSQL capabilities through MCP

Choose the connection path that fits the job. Your assistant inspects the available tools before configuring or running an action.

Sync management

Inspect tables and columns

Use connection schema discovery to identify real table names, column names and types before writing SQL or mapping a sync.

Connector actions

Prepare a bounded query

Discover the database query module and provide a connection, explicit columns, filters and a row limit. Free SQL queries require editor or owner access even for SELECT.

Sync management

Plan app write-back

Prepare a sync that propagates selected database changes to a supported app after reviewing direction, record matching and database permissions.

03 / Setup

Set up Stacksync MCP in Claude Code

Connect the MCP server first. Then connect PostgreSQL using the requirements below.

TerminalClaude Code

Add the remote Stacksync server to Claude Code and authenticate from its MCP menu.

Before you start

  • A Stacksync account with access to the workspace you want to use.
  • Permission to add an external MCP server in your AI client.
  • Claude Code installed with the claude command available in your terminal.
  1. Add Stacksync over HTTP

    Run the command below. Add --scope user if you want the connection available across your projects.

    claude mcp add --transport http stacksync https://mcp.stacksync.com/mcp

  2. Finish browser sign-in

    Open Claude Code, run /mcp, select stacksync and complete the Stacksync sign-in. Choose the workspace you want requests to use by default.

  3. Check your workspace

    Ask “List my Stacksync workspaces.” Confirm the workspace before choosing a connector or preparing a change. Connecting the assistant does not start a sync.

Check the connection

Ask “List my Stacksync workspaces.” Select the workspace you want to use, then ask the assistant to inspect the available PostgreSQL connection and capabilities.

Full setup documentation ↗

04 / Put it to work

PostgreSQL workflow prompts

Example prompts for PostgreSQL in Claude Code. The outcomes describe what to expect, not results from your account. Review the proposed configuration before running it.

01
Connector actions

Prepare a bounded operational query

Inspect my PostgreSQL connections and ask which database to use. Read the relevant table and column names, then propose a SELECT for the business question I provide, with explicit columns, a WHERE clause and LIMIT 100. Show the query and connection before execution. Explain the editor or owner requirement for the free-SQL action, even though the proposed SQL only reads data.

What to expect

A reviewable query grounded in the connected schema and scoped to a bounded result.

02
Sync management

Plan a CRM write-back table

Inspect my PostgreSQL and CRM connections and existing syncs. Propose how an approved score column could flow from a selected database table to the matching CRM record. Check primary keys, source identifiers, supported writes and any per-field direction settings I must configure in the web app. Do not create, start or change the sync yet.

What to expect

A write-back design with explicit IDs, direction and human setup requirements.

03
Sync management

Diagnose a missing sync table

Inspect my PostgreSQL connection and the tables exposed to Stacksync. Compare them with the table name I provide and review relevant sync issues. Propose checks for schema visibility, a single generated primary key, table permissions and change capture. Do not run permission changes, schema changes or database writes.

What to expect

A focused checklist tied to the missing table and the connection’s observed schema.

05 / Before you run

Requirements and limitations

What you need

  • Connect your AI client to Stacksync MCP, sign in and select the workspace before inspecting its resources.
  • Reuse an authorized app connection or open the secure connection link from Stacksync. Enter credentials in the connection form, never in the chat.
  • Ensure Stacksync can reach the database and the connection has the query or sync privileges required for the task. Synced tables need a single generated primary key.

What to account for

  • The documented connector action treats free SQL, including SELECT, as a change for permission and confirmation purposes. Read-only OAuth access is insufficient for that action.
  • Available actions depend on the live connector catalog, your connected account and MCP Gateway Policies. Inspect the action and its form before using it.
  • Logical replication or trigger-based change capture and network access require database setup. Per-field sync direction is configured in the Stacksync web app.

Your workspace. Your permissions.

OAuth access follows your Stacksync role. Choose read-only access during sign-in when appropriate. Workspace owners can enable “Ask before every change”; it is a configurable setting. Direct connector actions also follow gateway policies.

Read about permissions and approvals ↗

If setup does not work

The MCP server does not appear
Check your client’s custom MCP settings and any plan or administrator prerequisites. Reconnect using the setup instructions above.
Your workspace or app is missing
Verify the Stacksync account and workspace selected during OAuth, then inspect the workspace’s connections. Add missing app credentials through the secure connection flow.
An action is unavailable or denied
Inspect the current connector modules and required inputs. Check your workspace role, read-only consent, gateway policy, and the source app’s API permissions.

06 / Common questions

PostgreSQL MCP questions

Can Stacksync MCP run SQL against PostgreSQL?

The connector skill documents a database Query action. Discover the module and its form, always supply the connection, and inspect schema before constructing SQL. A free SQL query requires an editor or owner role.

Does read-only MCP consent permit a SELECT query?

The documented free-SQL action is classified as a change even for SELECT, because its input is arbitrary SQL. Read-only consent is suitable for inspection tools but does not grant this action.

What does a PostgreSQL table need for sync?

The connector documentation requires a single generated primary key and appropriate permissions. Network access and the chosen change-capture method must also be configured; ordinary SQL access alone does not establish sync readiness.