zektor

Command-line client for Zektor.io managed databases and caches.

Beta. The CLI and its MCP server are supported and live, but young. Expect commands to be added and rough edges filed off as we learn how people use them.

Beta is not a warning about stability of what exists: 1.0.0 settles the command surface. Flags, output shapes and exit codes will not change without a major version, so --json is safe to parse in a script, and tokens you create now keep working.

Install

npm install -g zektor

Or run it without installing:

npx zektor whoami

Requires Node 18 or newer.

Authenticating

Create a token in the dashboard under Settings → Access tokens, then:

zektor login

The token is stored at ~/.config/zektor/config.json with mode 0600. It is verified against the API before being written, so a mistyped token fails here rather than confusingly on your next command.

--token exists for scripts, but prefer the prompt when you are at a keyboard — an argument lands in your shell history.

In CI

Set ZEKTOR_TOKEN and skip login entirely:

ZEKTOR_TOKEN=zk_… zektor whoami

ZEKTOR_TOKEN always beats the config file. zektor whoami reports which of the two supplied the credential, which is usually how you discover a stray environment variable is overriding the account you logged in as.

Commands

Every read command takes --json.

Trying it without an account

psql "$(npx zektor try)"

zektor try prints a connection string on stdout and, on stderr, how long it works and a claim link. The database accepts connections for 60 minutes. Open the claim link within 3 hours, sign up and add a payment method, and the data moves into a database of your own; otherwise it is deleted. Up to 200 MB and 3 connections, no superuser (trusted extensions such as pgcrypto work), and 3 trials per network per day. --json prints all of it as JSON. The same is available without the CLI:

curl -X POST https://api.zektor.io/api/trial

Creating

zektor db create --name=my-db --tier=AKPG-5 --region=nbg1 --engine=postgres@17

--tier, --region and --engine take the names you see in the dashboard and are resolved to ids for you. --tier and --region are required; leave one out or pass a wrong one and the error lists what is valid.

Creating runs in the background. Add --wait to block until the instance is ready; if creating it fails, the command prints why and exits non-zero.

--storage is rejected: storage starts at what the plan includes. Grow it afterwards with zektor storage resize.

Connecting

connect opens redis-cli for a cache. For Postgres it prints the command shape instead — the password is shown once when a role is created or rotated and the server never stores it, so t

... View full README on GitHub