Yes. In Prefect 3.x you can tune the API server’s PostgreSQL connection pool via the built-in SQLAlchemy settings. The available knobs are: - pool_size - max_overflow - pool_timeout - pool_recycle You can set them with environment variables or in prefect.toml. Environment variables

# Required DB URL (example uses asyncpg)
PREFECT_SERVER_DATABASE_CONNECTION_URL="<postgresql+asyncpg://user:pass@host:5432/dbname>"

# Connection pool tuning
PREFECT_SERVER_DATABASE_SQLALCHEMY_POOL_SIZE=20
PREFECT_SERVER_DATABASE_SQLALCHEMY_MAX_OVERFLOW=40
PREFECT_SERVER_DATABASE_SQLALCHEMY_POOL_TIMEOUT=30
PREFECT_SERVER_DATABASE_SQLALCHEMY_POOL_RECYCLE=1800

prefect.toml

[server.database]
connection_url = "<postgresql+asyncpg://user:pass@host:5432/dbname>"

[server.database.sqlalchemy]
pool_size = 20
max_overflow = 40
pool_timeout = 30
pool_recycle = 1800

To confirm your current values at runtime:

prefect config view --show-defaults

Docs - Settings reference (including server/database nesting and env var names): Prefect 3 Settings reference - Server database/SQLAlchemy settings (pool_size, max_overflow, pool_timeout, pool_recycle): Server database configuration (API reference) Using PgBouncer - You can place PgBouncer in front of your Postgres instance; it’s a common way to scale beyond the DB’s native connection limits. Choose a pooling mode that matches your workload and driver; session pooling is the most compatible, while transaction pooling offers higher density but can affect session-level features. Prefect does not require session state, but your broader environment might. If you can share your target concurrency and database max_connections, I’m happy to suggest starting values for these settings.