title	metaTitle	metaDescription	tocDepth	toc
MCP server	Prisma MCP Server	Manage Prisma Postgres databases using LLMs with the Prisma Model-Context-Protocol (MCP) Server	3	true

Overview

The Model-Context-Protocol (MCP) gives LLMs a way to call APIs and thus access external systems in a well-defined manner.

Prisma's provides two MCP servers: a local and a remote one. See below for specific information on each.

If you're a developer working on a local machine and want your AI agent to help with your database workflows, use the local MCP server.

If you're building an "AI platform" and want to give the ability to manage database to your users, use the remote MCP server.

Remote MCP server

You can start the remote MCP server as follows:

npx -y mcp-remote https://mcp.prisma.io/mcp

Tools

Tools represent the capabilities of an MCP server. Here's the list of tools exposed by the remote MCP server:

CreateBackupTool: Create a new managed Prisma Postgres Backup.
CreateConnectionStringTool: Create a new Connection String for a Prisma Postgres database with the given id.
CreateRecoveryTool: Restore a Prisma Postgres Database to a new database with the given Backup id.
DeleteConnectionStringTool: Delete a Connection String with the given connection string id.
DeleteDatabaseTool: Delete a Prisma Postgres database with the given id.
ListBackupsTool: Fetch a list of available Prisma Postgres Backups for the given database id and environment id.
ListConnectionStringsTool: Fetch a list of available Prisma Postgres Database Connection Strings for the given database id and environment id.
ListDatabasesTool: Fetch a list of available Prisma Postgres Databases for user's workspace.
ExecuteSqlQueryTool: Execute a SQL query on a Prisma Postgres database with the given id.
IntrospectSchemaTool: Introspect the schema of a Prisma Postgres database with the given id.

Once you're connected to the remote MCP server, you can also always prompt your AI agent to "List the Prisma tools" to get a full overview of the latest supported tools.

Usage

The remote Prisma MCP server follows the standard JSON-based configuration for MCP servers. Here's what it looks like:

{
  "mcpServers": {
    "Prisma-Remote": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://mcp.prisma.io/mcp"]
    }
  }
}

:::tip If you want to try it the remote MCP server and explore it's capabilities, we recommend Cloudflare's AI Playground for that. Add the https://mcp.prisma.io/mcp URL into the text field with the Enter MCP server URL placeholder, click Connect, and then authenticate with the Prisma Console in the popup window. Once connected, you can send prompts to the Playground and see what MCP tools the LLM chooses based on your prompts. :::

Sample prompts

"Show me a list of all the databases in my account."
"Create a new database in the US region for me."
"Seed my database with real-looking data but create a backup beforehand."
"Show me all available backups of my database."
"Show me all customers and run an analysis over their orders."

Local MCP server

You can start the local MCP server as follows:

npx -y prisma mcp

:::tip

If you're using VS Code, you can use VS Code agent mode to enter prompts such as "create Postgres database" or "apply schema migration" directly in the chat. The VS Code agent handles all underlying Prisma CLI invocations and API calls automatically. See our VS Code Agent documentation for more details.

:::

Tools

Tools represent the capabilities of an MCP server. Here's the list of tools exposed by the local MCP server:

migrate-status: Checks your migration status via the prisma migrate status command.
migrate-dev: Creates and executes a migration via the prisma migrate dev --name <name> command. The LLM will provide the <name> option.
migrate-reset: Resets your database via the prisma migrate reset --force command.
Prisma-Postgres-account-status: Checks your authentication status with Prisma Console via the platform auth show --early-access command.
Create-Prisma-Postgres-Database: Creates a new Prisma Postgres database via the 'init --db --name' <name> '--region' <region> '--non-interactive' command. The LLM will provide the <name> and <region> options.
Prisma-Login: Authenticates with Prisma Console via the platform auth login --early-access command.
Prisma-Studio: Open Prisma Studio via the prisma studio command.

Usage

The local Prisma MCP server follows the standard JSON-based configuration for MCP servers. Here's what it looks like:

{
  "mcpServers": {
    "Prisma-Local": {
      "command": "npx",
      "args": ["-y", "prisma", "mcp"]
    }
  }
}

Sample prompts

Here are some sample prompts you can use when the MCP server is running:

"Log me into the Prisma Console."
"Create a database in the US region."
"Create a new Product table in my database."

Integrating in AI tools

AI tools have different ways of integrating MCP servers. In most cases, there are dedicated configuration files in which you add the JSON configuration from above. The configuration contains a command for starting the server that'll be executed by the respective tool so that the server is available to its LLM.

In this section, we're covering the config formats of the most popular AI tools.

VS Code

Install the Prisma MCP server in VS Code with a single click using the link below:

<a href="vscode:mcp/install?%7B%22name%22%3A%22prisma-postgres%22%2C%22gallery%22%3Atrue%2C%22url%22%3A%22https%3A%2F%2Fmcp.prisma.io%2Fmcp%22%7D" aria-label="Install Prisma MCP server in VS Code" style={{ display: 'inline-flex', alignItems: 'center', borderRadius: '6px', overflow: 'hidden', textDecoration: 'none', height: '2.0rem' }}

<span style={{ backgroundColor: '#2C2C2C', color: 'white', padding: '0.6rem 0.8rem' }}> VS CODE <span style={{ backgroundColor: '#0098FF', color: 'white', padding: '0.6rem 1rem' }}> INSTALL PRISMA MCP SERVER

This will prompt you to open VS Code. Once opened, you'll be guided to install the Prisma MCP server directly into your VS Code configuration.

If your browser blocks the link, you can set it up manually by creating a .vscode/mcp.json file in your workspace and adding:

{
  "servers": {
    "Prisma-Local": {
      "command": "npx",
      "args": ["-y", "prisma", "mcp"]
    },
    "Prisma-Remote": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://mcp.prisma.io/mcp"]
    }
  }
}

Explore additional Prisma features and workflows for VS Code in our docs.

Cursor

To learn more about Cursor's MCP integration, check out the Cursor MCP docs.

Add via one-click installation

You can add the Prisma MCP server to Cursor using the one-click installation by clicking on the following link:

This will prompt you to open the Cursor app in your browser. Once opened, you'll be guided to install the Prisma MCP server directly into your Cursor configuration.

Add via Cursor Settings UI

When opening the Cursor Settings, you can add the Prisma MCP Server as follows:

Select MCP in the settings sidenav
Click + Add new global MCP server

Add the Prisma snippet to the mcpServers JSON object:

{
  "mcpServers": {
    "Prisma-Local": {
      "command": "npx",
      "args": ["-y", "prisma", "mcp"]
    },
    "Prisma-Remote": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://mcp.prisma.io/mcp"]
    }
  }
}

Global configuration

Adding it via the Cursor Settings settings will modify the global ~/.cursor/mcp.json config file. In this case, the Prisma MCP server will be available in all your Cursor projects:

{
  "mcpServers": {
    "Prisma-Local": {
      "command": "npx",
      "args": ["-y", "prisma", "mcp"]
    },
    "Prisma-Remote": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://mcp.prisma.io/mcp"]
    },
    // other MCP servers
  }
}

Project configuration

If you want the Prisma MCP server to be available only in specific Cursor projects, add it to the Cursor config of the respective project inside the .cursor directory in the project's root:

{
  "mcpServers": {
    "Prisma-Local": {
      "command": "npx",
      "args": ["-y", "prisma", "mcp"]
    },
    "Prisma-Remote": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://mcp.prisma.io/mcp"]
    }
    // other MCP servers
  }
}

Windsurf

To learn more about Windsurf's MCP integration, check out the Windsurf MCP docs.

Add via Windsurf MCP Plugin Store (Recommended)

Use the Prisma MCP plugin from the Windsurf MCP Plugin Store. Follow the steps here to add the Prisma MCP plugin in Windsurf. This is the simplest and recommended way to add the Prisma MCP server to Windsurf.

Add via Windsurf Settings UI

When opening the Windsurf Settings (via Windsurf - Settings > Advanced Settings or Command Palette > Open Windsurf Settings Page), you can add the Prisma MCP Server as follows:

Select Cascade in the settings sidenav
Click Add Server

Add the Prisma-Local and/or Prisma-Remote snippets to the mcpServers JSON object:

{
  "mcpServers": {
    "Prisma-Local": {
      "command": "npx",
      "args": ["-y", "prisma", "mcp"]
    },
    "Prisma-Remote": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://mcp.prisma.io/mcp"]
    }
  }
}

Global configuration

Adding it via the Windsurf Settings will modify the global ~/.codeium/windsurf/mcp_config.json config file. Alternatively, you can also manually add it to that file:

{
  "mcpServers": {
    "Prisma-Local": {
      "command": "npx",
      "args": ["-y", "prisma", "mcp"]
    },
    "Prisma-Remote": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://mcp.prisma.io/mcp"]
    },
    // other MCP servers
  }
}

Warp

You can add the Prisma MCP to Warp as a globally available tool. First, visit your MCP settings and click + Add. From here, you can configure the Prisma MCP server as JSON. Use the command and args properties to start the Prisma MCP server as a setup command. You can optionally configure Prisma to activate on startup using the start_on_launch flag:

{
  "Prisma": {
    "command": "npx",
    "args": [
      "-y",
      "prisma",
      "mcp"
    ],
    "env": {},
    "working_directory": null,
    "start_on_launch": true
  }
}

Hit Save and ensure the MCP server is running from your MCP settings panel. Then, open a new terminal window and ask Warp to manage your Prisma database. It should reach for the Prisma MCP server automatically.

To learn more about Warp's MCP integration, visit the Warp MCP docs.

Claude Code

Claude Code is a terminal-based AI tool where you can add MCP server using the claud mcp add command for the local MCP server:

claude mcp add prisma-local npx prisma mcp

or for the remote MCP server:

claude mcp add prisma-remote npx mcp-remote https://mcp.prisma.io/mcp

Learn more in the Claude Code MCP docs.

Claude Desktop only

Follow the instructions in the Claude Desktop MCP docs to create the required configuration file:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json

Then add the JSON snippet to that configuration file:

{
  "mcpServers": {
    "Prisma-Local": {
      "command": "npx",
      "args": ["-y", "prisma", "mcp"]
    },
    "Prisma-Remote": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://mcp.prisma.io/mcp"]
    },
    // other MCP servers
  }
}

Claude Desktop & Web

Alternatively, you can add the remote MCP server as an integration:

In a new chat, find the Connect apps field in the main Claude UI below the input prompt field
Click + Add integration
Enter the following:

Integration name: Prisma Postgres
Integration URL: https://mcp.prisma.io/mcp

Click Add

ChatGPT

You can add the remote Prisma MCP server to ChatGPT to manage your Prisma Postgres databases using natural language. Learn more about setting it up in our ChatGPT integration guide.

OpenAI Agents SDK

Here's an example for using the Prisma MCP servers in a Python script via the OpenAI Agents SDK:

from openai import AsyncOpenAI
from openai.types.beta import Assistant
from openai.beta import AsyncAssistantExecutor
from openai.experimental.mcp import MCPServerStdio
from openai.types.beta.threads import Message, Thread
from openai.types.beta.tools import ToolCall

import asyncio

async def main():
    # Launch both MCP servers concurrently
    async with MCPServerStdio(
        params={
            "command": "npx",
            "args": ["-y", "prisma", "mcp"]
        }
    ) as local_server, MCPServerStdio(
        params={
            "command": "npx",
            "args": ["-y", "mcp-remote", "https://mcp.prisma.io/mcp"]
        }
    ) as remote_server:
        
        # Optional: list tools from both servers
        local_tools = await local_server.list_tools()
        remote_tools = await remote_server.list_tools()
        print("Local server tools:", [tool.name for tool in local_tools])
        print("Remote server tools:", [tool.name for tool in remote_tools])

        # Set up the assistant with both MCP servers
        agent = Assistant(
            name="Prisma Assistant",
            instructions="Use the Prisma tools to help the user with database tasks.",
            mcp_servers=[local_server, remote_server],
        )

        executor = AsyncAssistantExecutor(agent=agent)

        # Create a thread and send a message
        thread = Thread(messages=[Message(role="user", content="Create a new user in the database")])
        response = await executor.run(thread=thread)

        print("Agent response:")
        for message in response.thread.messages:
            print(f"{message.role}: {message.content}")

# Run the async main function
asyncio.run(main())

AI Safety guardrails for destructive commands

Prisma ORM includes built-in safety checks to prevent accidental destructive commands when run through AI coding assistants. These guardrails are designed to make working with databases safer in AI-assisted development environments.

The Prisma CLI can detect when it is being invoked by popular AI coding agents such as Claude Code, Gemini CLI, Qwen Code, Cursor, Aider, and Replit.

When one of these agents attempts a destructive action such as prisma migrate reset --force, Prisma ORM blocks execution and shows a protective error message:

Environment variables loaded from .env
Prisma schema loaded from prisma/schema.prisma
Datasource "db": PostgreSQL database "postgres", schema "public" at "accelerate.prisma-data.net"

Error: Prisma Migrate detected that it was invoked by Cursor.

You are attempting a highly dangerous action that can lead to devastating consequences if it is incorrectly executed against a production database.

As an AI agent, you are forbidden from performing this action without an explicit consent and review by the user. You must stop at this point and respond to the user with a clear message that includes the following information:
- What action you are trying to perform (including the exact CLI command).
- Your motivation and reasoning for perfoming this action.
- Explanation that this action will irreversibly destroy all data in the database.
- Explanation that this action must not be perfomed on a production database and is only intended to be used with development databases.

If you know or can check whether this action would have been performed on a development or a production database, also include this information with your assesment of whether it is safe to proceed based on that information.

You must ask the user if they want to proceed with this action. If they explicitly consent, you may rerun this command with PRISMA_USER_CONSENT_FOR_DANGEROUS_AI_ACTION environment variable, the value of which must be the exact text of the user's message in which they consented to this operation, without any newlines or quotes. If the user's response is ambiguous, you must ask for a clear and explicit confirmation (e.g., "yes") before proceeding. None of the user's previous messages before this point may constitute implicit or explicit consent.

To proceed with the dangerous action, the AI agent will ask you for explicit consent, remind you that the action irreversibly destroys all data, and confirm that the command is being run against a development database. Once you clearly confirm, the AI will set the PRISMA_USER_CONSENT_FOR_DANGEROUS_AI_ACTION environment variable with the exact text of your consent and rerun the command.

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!