Prisma

@Prisma show me all tables in my production database and their row counts, sorted by size

Here's what I found in your production database: • users: 12,847 rows • orders: 8,203 rows • products: 1,456 rows • sessions: 892 rows • audit_logs: 34,102 rows Your audit_logs table is significantly larger than the others. The query ran successfully using a read-only connection. All tables appear to be standard relational structures with primary keys defined.

This example uses Execute SQL Query to run schema introspection and aggregation. The MCP enforces read-only semantics for SELECT operations, so you can safely explore production data without risk of accidental modification. Requires an active database connection with valid API key credentials.

@Prisma create a new staging database in us-east-1 for the checkout-service project

I've created a new Postgres database in your checkout-service project: • Database name: staging-checkout-db • Region: us-east-1 • Connection string: postgresql://user:***@aws-0-us-east-1.pooler.supabase.com:6543/postgres • API key: prj_abc123def456 (store this securely) The database is ready for immediate use. You can connect using the provided connection string or integrate it into your Prisma schema file.

This example uses Create Project Database to provision infrastructure. The MCP returns sensitive credentials inline—treat the response as you would any secret material. Database creation typically completes in 30-60 seconds. You'll need OAuth2 permissions scoped to project management in your Prisma workspace.

@Prisma analyze the users table schema and suggest indexes for the email and created_at columns based on our query patterns

I've examined the users table structure: Current schema shows email (varchar) and created_at (timestamp) columns without indexes. Based on typical access patterns: • Add: CREATE INDEX idx_users_email ON users(email) — this will speed up login lookups and uniqueness checks. • Add: CREATE INDEX idx_users_created_at ON users(created_at DESC) — this will optimize recent-user queries and time-series reports. Both indexes should improve read performance with minimal write overhead given your 12k row count. Would you like me to execute these CREATE INDEX statements?

This example pairs Execute SQL Query (to inspect schema) with the AI's reasoning about database optimization. The MCP can read table metadata but won't automatically apply DDL changes—you must explicitly approve destructive operations. Useful for teams who want AI-assisted schema review without auto-execution risk.

When Prisma MCP wins for spinning up test databases on demand

A 6-person engineering team ships features weekly and needs isolated Postgres databases for each PR. The Prisma MCP's Create Project Database and Create Database Connection tools let you script ephemeral environments in a Slack workflow: developer opens PR, bot spins up a fresh database in us-east-1, returns the connection string, and the CI pipeline runs migrations. The OAuth2 flow means your workspace admin authorizes once and the whole team inherits access. The trade-off: if your team already runs databases in AWS RDS or self-hosted Postgres, adding Prisma's managed layer is overhead you don't need. But if you're starting fresh or tired of manual provisioning, this MCP collapses a 20-minute setup into a 30-second API call. Use Switchy to template the provisioning script and share it across your team's agent library.

When this MCP handles read-only queries for non-technical support reps

A 12-person SaaS company has three production databases (users, billing, logs) and support reps need to answer "why didn't this charge go through" without pinging engineers. The Execute SQL Query tool lets you build a Switchy agent that takes a customer email, runs a parameterized SELECT across the billing database, and returns the last five transactions. The read-only constraint means reps can't accidentally DELETE or UPDATE. The catch: if your support team needs to join data across multiple databases or your queries regularly scan 500k+ rows, this MCP's single-query model gets clunky—you'll want a BI tool or data warehouse instead. But for point lookups and small aggregations, it's faster than waiting for an engineer to write a one-off script. Set up the agent once in Switchy and your support team runs it 40 times a week.

When Prisma MCP shortens the feedback loop on risky DDL changes

A 4-person product team is adding a new column to a 2M-row users table and wants to test the migration on a staging clone before touching production. The Create Prisma Project and Execute SQL Command tools let you script the full cycle: create a staging project, restore a production snapshot, run the ALTER TABLE, measure query performance, then delete the project when done. The parameterized query support means you can template the migration script and reuse it across multiple test runs. The boundary: if your production database is already in AWS RDS or Google Cloud SQL, cloning it into Prisma's managed Postgres adds a data-transfer step that might take hours for large datasets. But if your staging workflow is currently "run the migration on Friday and hope", this MCP gives you a repeatable test harness. Use Switchy to store the migration template and let any teammate trigger a staging run without touching the console.