Connect Prisma Postgres to Excel, Google Sheets, AI and REST

Two connection strings, and only one of them is Postgres

This is the thing that catches people, and it catches them because the wrong string looks perfectly reasonable.

• Prisma Accelerate URLs start with prisma:// or prisma+postgres://. They address Prisma's own connection layer, which speaks its protocol rather than the Postgres wire protocol.
• Query Streams connects over direct TCP with a standard Postgres driver, so an Accelerate URL cannot work here no matter what credentials it carries.
• Paste one anyway and the card tells you exactly that, and what to generate instead. It is a specific message rather than a generic parse failure, because the two strings are easy to mix up.
• The one you want starts with postgres:// or postgresql:// and points at db.prisma.io on 5432.

Everyone shares one hostname

Most managed Postgres services give you a hostname unique to your instance. Prisma does not, and it changes what the fields mean.

• Every Prisma Postgres database answers on db.prisma.io. There is no per-project or per-region hostname to look up.
• Your username and password are what select the database. They are generated in the Prisma Console rather than chosen by you, and they are the entire identity of the connection.
• That makes the credentials worth treating carefully. On most services a leaked password is scoped by an obscure hostname as well; here the hostname is public knowledge.
• The pasted string is never stored. It is read once to fill the form and then dropped, so the password lives in one place rather than two.

TLS is required and the agent does not leave it to chance

Because the hostname is fixed and known, the agent can be firm about this one.

• The agent forces TLS for db.prisma.io. The Use SSL checkbox defaults on, and clearing it does not produce an unencrypted connection to a Prisma host.
• The certificate chain is not verified. Encryption is required; validation of what signed the certificate is not performed, so read it as protection against eavesdropping rather than proof of what you reached.
• There is no connection pooler on this path. Prisma's pooling story lives in Accelerate, which this card deliberately does not use, so what you get is a direct connection with no pooled mode to choose.
• Everything downstream is ordinary Postgres: schemas, types, and public as the default schema in a cross-source reference.

-- Ordinary Postgres, read-only, over direct TCP
SELECT   c.name,
         COUNT(o.id)   AS orders,
         SUM(o.amount) AS revenue
FROM     public.customers AS c
JOIN     public.orders    AS o ON o.customer_id = c.id
WHERE    o.placed_at >= now() - interval '30 days'
GROUP BY c.name
ORDER BY revenue DESC;