Remote MCP endpoint
Meet ChatGPT's requirement for a reachable remote MCP server without operating your own transport.
ChatGPT / PostgreSQL / MCP app
Publish a remote PostgreSQL MCP app for ChatGPT with OAuth, explicit database permissions, encrypted credentials, and a record of every query. Developer mode must be available for the account or workspace.
tools/call → productionPOLICY CHECKEDThe production path
ChatGPT can use remote MCP apps to discover tools and act on connected systems. DataMCP turns PostgreSQL schema inspection and approved SQL operations into those tools, then hosts the remote endpoint ChatGPT requires.
The gateway separates ChatGPT from the database secret. Users authorize through OAuth, while each MCP link defines whether ChatGPT can only read data, write approved records, or access selected tables.
For the current ChatGPT controls, read the ChatGPT MCP setup guide. Then follow the PostgreSQL-specific walkthrough for database permissions and testing.
What you get
Meet ChatGPT's requirement for a reachable remote MCP server without operating your own transport.
Users authorize the app without copying PostgreSQL credentials or shared database secrets into ChatGPT.
Ask business questions while ChatGPT inspects the live schema and executes approved SQL tools.
Start read-only or choose write and custom table permissions when the workspace and use case require them.
PostgreSQL credentials remain encrypted on DataMCP and are never exposed to the ChatGPT app.
Review the database calls made through the MCP app, including time, user context, status, and result metadata.
Source to controlled MCP link
Connect the source once. DataMCP handles discovery, hosted transport, client authentication, and link policy.
Read the setup guideAdd the database to DataMCP and verify the discovered schema and available database tools.
Choose a safe permission preset, then copy the ChatGPT-compatible remote endpoint.
Enable Developer mode, add the endpoint from Settings → Plugins, complete OAuth, review tools, and test the app.
Build or use datamcp
See what stays on your engineering backlog when the MCP runtime and control layer are managed for you.
Custom code or unsupported database socket
Shared PostgreSQL credential
Whatever the database user can do
Distributed to integration code
Assemble from application and database logs
Frequently asked questions
Clear answers about architecture, access, credentials, and supported clients.
ChatGPT does not connect to a PostgreSQL socket directly. It connects to a remote MCP app that exposes database tools. DataMCP provides that remote endpoint and handles PostgreSQL connectivity, authentication, permissions, and query execution.
OpenAI currently says ChatGPT Apps are supported on all plans. Developer mode can still be unavailable when an organization administrator disables it, and workspace policy controls app permissions.
No. ChatGPT authenticates to the DataMCP endpoint through OAuth. The PostgreSQL connection string remains encrypted on DataMCP and is not provided to ChatGPT or stored in the app configuration.
Yes. Use a read-only MCP link for analytics and exploration, or define a custom table-level policy. ChatGPT can only call database tools allowed by both the DataMCP link and the underlying PostgreSQL role.
Only when the MCP link and workspace allow write actions. OpenAI currently describes full MCP write support as a beta feature for eligible workspaces, and ChatGPT may request confirmation before consequential actions.
Keep exploring
Use the current Developer mode, Plugins, permissions, testing, and refresh workflow.
GUIDEWalk through database connection, MCP permissions, OAuth, and the first query.
GUIDEUnderstand the database tools, permission model, and hosted architecture behind the app.
GUIDESee how authentication, secrets, policy checks, and audit logs fit together.
Start with one link
Connect one PostgreSQL database and create one controlled remote MCP link for free. Add it where ChatGPT Developer mode is available.