Build and inspect database connection strings for six engines, with credentials masked by default and revealing made a deliberate, local-only action.
PostgreSQL connection string
The masked form is what copy takes until you reveal.
A connection string is a credential. Nothing you type here is uploaded, logged or stored — but anything on screen is visible in a screen share, which is why values are masked until you ask.
A URL. `postgres://` and `postgresql://` are both accepted everywhere; libpq treats them as identical.
postgresql://app@localhost:5432/app_development
No `sslmode`. libpq defaults to `prefer`, which uses TLS if the server offers it and silently falls back to plaintext if it does not — so an unencrypted connection looks exactly like an encrypted one. `verify-full` is the setting that fails loudly instead.
Keep this in an environment variable rather than in a file you commit. The env form below is what belongs in your secrets manager.
Connection String Builder runs entirely on your device — no connection string is uploaded, logged or stored, and no database is contacted.
You’re on the Free plan. Compare plans
A connection string is a credential. It carries a username, a password, an internal hostname and often a port that is not meant to be public — which makes this the most sensitive input in the catalogue after a private key, and shapes every decision in the tool.
Everything on screen is masked by default. Revealing is a deliberate click held in local state and never persisted. Nothing is uploaded, nothing is logged, and nothing is written to browser storage of any kind — not localStorage, not sessionStorage, not IndexedDB.
The copy button takes the masked form until you reveal, which means a copy made during a screen share cannot carry a password out by accident. The environment-variable form necessarily contains the real password — it is the line you paste into a `.env` file — so it appears only once you have revealed, rather than sitting beside the masked string where a screenshot would catch it.
The inspector has no reveal control at all. Knowing that a password is present and how long it is answers the question you came with, without putting a live credential on screen.
Each engine's quirks are explained where they bite: MongoDB's `+srv` scheme does a DNS lookup for the host list and forbids a port, PostgreSQL's `sslmode=require` encrypts without verifying the certificate, and a password containing `@` or `/` has to be percent-encoded or the string parses into the wrong parts entirely.
Nothing is uploaded, logged, or written to browser storage, and the analytics schema has no field that could carry a value. The remaining risk is your own screen, which is why the password is masked until you ask and the inspector has no reveal control at all.
Almost certainly a character that means something in a URL. An `@` separates credentials from the host, a `/` starts the path and a `:` separates user from password — each has to be percent-encoded or the string parses into the wrong parts.
No, and the name is misleading. `require` encrypts the connection without checking who is at the other end, so it stops passive eavesdropping and not an active attacker. `verify-full` is the one that checks the certificate and the hostname.
It makes the driver look up the host list in DNS rather than reading it from the string, which is how replica set membership can change without a redeploy. It forbids an explicit port, because the SRV records carry them.
In an environment variable, never in source. The tool emits that form for you — and if a connection string has ever been committed, rotate the password rather than just removing the line, because the history keeps it.