Get Started
Pick the surface that fits your situation, and ship your first note in five minutes.
There are two ways to use Plext today. They share the same data model, the same authorization rules, and the same notes — pick whichever matches your situation.
I'm a developer at my terminal
Use the CLI.
# Install (Homebrew, npm, or curl-pipe — pick one)
brew install plext-cli
# Authenticate, link a workspace, install agent skills — one command
plext setupplext setup opens an OAuth browser flow, lets you pick which plexes to capture into, and installs the plext-write and plext-read skills into Claude Code, Codex CLI, or Cursor (whichever you have). Once installed, your agent can save and recall knowledge as you work.
I'm building a backend that talks to Plext
Use the API.
export PLEXT_KEY="plxt_***REDACTED***"
export PLEXT_BASE="https://api.plext.com/v1"
# Verify the key
curl -s "$PLEXT_BASE/me" -H "Authorization: Bearer $PLEXT_KEY" | jqAPI keys are minted from the workspace settings page on plext.com. They're scoped to specific plexes and capabilities at mint time, long-lived (default 1 year), and rotatable.
What you'll need either way
- A Plext account — sign up at plext.com.
- A workspace — created automatically when you sign up. You can be a member of many.
- At least one plex — every workspace ships with three plexes (
general,bug-reports,tool-evaluations). Create more as needed.
Not sure which to pick?
| Scenario | Use |
|---|---|
| AI coding agent inside Claude Code / Cursor / Codex CLI | CLI (installs the skill into the agent) |
| Human running terminal commands | CLI |
| Your own server, Lambda, Worker, cron job | API |
| CI/CD saving release notes | API (or CLI if the runner is logged in) |
| Custom dashboard reading plex/note state | API |
| Background worker pulling from Slack/email into notes | API |
Rule of thumb: humans use the CLI; servers use the API. AI agents in IDEs are reached via the CLI, which installs skills that call the API under the hood.