ModuleNotFoundError: No module named 'ipykernel' in jupyter lab, how to fix

TL;DR

JupyterLab is running fine, but the python environment behind your kernel never got ipykernel installed, so no kernel process can start. Run python -m pip install ipykernel in that exact environment, then python -m ipykernel install --user --name myenv --display-name "Python (myenv)". The kernel then appears in the launcher and connects on the first try.

ModuleNotFoundError: No module named 'ipykernel'

Use this when

• JupyterLab opens but starting any python kernel fails with ModuleNotFoundError for ipykernel.
• The traceback or server log shows the missing module right after kernel launch is attempted.
• A fresh conda env or venv shows up in kernelspec list but cannot start.

Not for this skill when

• The kernel starts and then dies mid-cell. That is a crash, not a missing package.
• JupyterLab itself will not launch. That is a server-side install problem.

Steps

1. Find which python the kernel spec uses: jupyter kernelspec list, then read the argv in the spec's kernel.json. Verify: you know the exact python path.
2. Install ipykernel into that python: [that python] -m pip install ipykernel. Verify: [that python] -c "import ipykernel" succeeds.
3. Register the spec: [that python] -m ipykernel install --user --name [env-name] --display-name "Python ([env-name])". Verify: jupyter kernelspec list shows the entry.
4. Restart JupyterLab and pick the new kernel in the launcher. Verify: the kernel status dot goes solid and 1+1 executes.
5. If the env is conda-managed, prefer conda install -n [env-name] ipykernel to keep the solver happy. Verify: conda list -n [env-name] ipykernel shows it.

Variant phrasings

no module named ipykernel jupyter

The generic search. Same missing-package fix.

ipykernel not installed kernel error

Log-style phrasing. Install into the kernel's python, not the server's.

jupyter lab kernel not found ipykernel

Launcher-side phrasing. The registration step is what makes it appear.

Compatibility: JupyterLab 3/4, notebook 6/7, ipykernel 6.x. The install-and-register flow is the same on pip and conda.

Why it happens

JupyterLab and the kernel are separate processes, often from separate environments. The Lab server only needs jupyter_server; the kernel needs ipykernel. Fresh virtualenvs and conda envs routinely get a kernelspec copied in without ipykernel ever being installed there, so Lab offers a kernel it cannot launch.

Edge cases / pitfalls

• Installing ipykernel into the Lab server's environment does not fix a kernel in another env. Match the python exactly.
• Do not edit kernel.json by hand to point at a different python. Reinstall the spec from the right python instead.
• --user installs go to your home kernels dir. In multi-user JupyterHub, register system-wide or per user deliberately.

Provenance

Resolved from the public thread: https://vectle.com/posts/pstUGKVg6h107NbMy6arNRpA