Install the free PostgreSQL
extension for VS Code

Why the editor

Your whole PostgreSQL workflow, in the editor you already use

• Stay in flow

  Connect, query, and manage PostgreSQL where you already write code.

• Write SQL that matches your schema

  IntelliSense knows your connected database, so names and completions are right the first time.

• Bring AI in on your terms

  Ask @pgsql about your data. Agent mode can act, but only with your approval.

• Diagnose and resolve issues

  Open the Metrics Dashboard and Azure Advisor recommendations without switching tools.

Hands-on

Build a research agent
backed by PostgreSQL on Azure

Turn what you learned into a working agent: PostgreSQL as the vector store, a Microsoft Foundry agent as the brain, and real Washington State case law as the data.

1. 01

   Step 1: Set up your environment

   Clone the repo, create a Python virtual environment, and add your keys to .env.

   Terminal

   $ git clone https://github.com/Azure-Samples/postgres-agents$ cd postgres-agents$ pip install -r requirements.txt

2. 02

   Step 2: Enable the extensions

   Allow-list azure_ai and vector on your Azure Database for PostgreSQL server, then create them.

   SQL

   CREATE EXTENSION IF NOT EXISTS azure_ai;CREATE EXTENSION IF NOT EXISTS vector;

3. 03

   Step 3: Load data and generate embeddings

   Create the cases table and embed the case opinions with Azure OpenAI. About 3–5 minutes.

   Terminal

   $ cd load_data$ python main.py$ cd ..

4. 04

   Step 4: Give your agent a Postgres tool

   vector_search_cases in legal_agent_tools.py runs a pgvector similarity search the agent can call.

   legal_agent_tools.py
   def vector_search_cases(query)
5. 05

   Step 5: Create and run the agent

   Initialize the agent in your Microsoft Foundry project with gpt-4o-mini, and answer a real legal question end to end.

   Terminal

   $ python src/simple_postgres_and_ai_agent.py

6. 06

   Step 6: Test and iterate

   Ask follow-up questions in the Microsoft Foundry playground and trace each step.

   PlanCall toolAnswer

Keep going

• Read the docs on Microsoft Learn (opens in a new tab)

  Overview, connections and identity, query editor, Copilot integration, @pgsql chat participant, Agent mode, MCP server.

• Explore more samples on the PostgreSQL Hub (opens in a new tab)

  Learning pathways, solution accelerators, and community.

• Rewatch the series on YouTube (opens in a new tab)

  Full playlist on Microsoft Developer.

• Get the VS Code ebook (opens in a new tab)

  A guide to working with PostgreSQL in VS Code.