Turn Jensen on

Installing Jensen put the app and the jensen command on your machine. Turning it on for a project is a separate step, and the app walks you through it.

The guided wizard

Open a project that has not been set up and the wizard starts on its own. You can reach it again later from Settings, System, Setup.

Nine steps. Five need a decision from you.

Step	What you decide	Required
Welcome	Nothing. It explains what Jensen builds for this project.	No
Import configuration	Whether to bring VS Code or Kiro configuration across. Your original files are left untouched.	No
Connect	Nothing. It reconciles hooks, the context server, the git guard and the index, reports how many are wired, and offers Fix all.	Yes
Appearance	Theme and interface size.	No
AI connection	Which assistant or API Jensen calls.	Yes
Git workflow	Which branch serves each environment.	Yes
Agent hooks	Which built-in agent hooks are on.	Yes
Project tools	Which language tools this project uses.	Yes
Complete	Nothing. Where to go next.	No

A separate flow drafts the project’s five context documents with AI. The draft stays local and editable, and nothing is saved until you review it. See Project context.

From the terminal

jensen setup

Same steps, no window.

What that wires

Step	What it does
cli-path	Links jensen onto your PATH so you can open any project from a terminal.
project-activate	Turns Jensen on for this project.
git-shim	Installs the git guard.
git-hooks	Points this project’s git hooks at the guard, so it runs on every commit however it was made.
mcp	Registers the context server once at user scope with every supported assistant found on your PATH.
hooks	Wires each assistant’s own session lifecycle hooks.
attribution	Silences the assistant’s commit byline.
debrief-command	Installs the /jensen-debrief command that saves what a session learned.
guidance	Writes the shared agent conventions into the project.
daemon	Checks the background service.
project	Indexes the project so its map and knowledge are ready to query.
state	Records what was wired.

It works out the answers itself

You do not describe your machine to Jensen. Setup goes and looks, and every decision it makes is scoped to the project it ran in.

• It finds your assistants. Whichever assistant command-line tools are installed get the context server registered, once at user scope, so you approve it a single time. It does not ask you to name them, and it does not install one.
• It wires each assistant the way that assistant expects. Session hooks, the plan format, the debrief command and the commit-byline setting are applied in that assistant’s own configuration.
• It works out your language tooling from the repository. See Language tools.
• It reconciles instead of overwriting. Every step probes what is already in place first. Setup never replaces configuration you own, and never touches hooks or binaries it did not install.
• It leaves the project’s settings in the project. Branch policy, agent hooks, tool overrides and project context are per project, so two repositories on the same machine can work completely differently.

Re-running it is safe

Every step checks what is already in place and leaves it alone.

From the terminal

jensen setup --status          # report what is wired, write nothing
jensen setup --no-scan         # skip indexing, the slow part on a large repo
jensen setup --only git-shim   # run one step
jensen setup --skip mcp        # run everything but one step
jensen setup --uninstall       # reverse every step, or the ones --only and --skip select

Finishing up

Restart your shell, or source your profile, so the newly linked command is found.

One assistant needs a one-time trust review for user-installed command hooks. Start Codex and use /hooks to review and trust the Jensen entries. Claude Code and Gemini CLI pick up their updated settings on the next session.

You are ready for Your first session.

If you would rather never open the app, everything above still applies. See Working outside the app.