Why does pip install fail with "No module named 'numpy'" when numpy is installed?

pip uses build isolation by default, which builds the package in a fresh environment containing only the packages listed in `[build-system].requires` in `pyproject.toml`. If `numpy` is needed at build time (to compile C extensions) but not listed there, the build environment cannot see the numpy installed in your project, and the build fails.

How do I fix the numpy build error during pip install?

Pass `--no-build-isolation` to use the current environment instead of an isolated one: `pip install --no-build-isolation .` or `uv pip install --no-build-isolation git+https://github.com/example/pkg.git`. This makes the already-installed numpy visible to the build. The flag works for one-off installs but trades reproducibility for convenience.

How do I permanently fix missing build dependencies in pyproject.toml?

Declare every build-time dependency in the `[build-system].requires` table. For a package that needs numpy headers to compile, set `requires = ["setuptools", "wheel", "numpy"]` with `build-backend = "setuptools.build_meta"`. This follows PEP 518, which specifies how a project declares its build-time requirements, and lets pip resolve the build environment correctly without `--no-build-isolation`.

How to Fix ModuleNotFoundError: No module named 'numpy' When Building a Package with pip

by Tim Hopper · Markdown

Scientific Python verified with pip 26.2 verified with uv 0.12

This guide helps resolve an issue where a Python package fails to install because it cannot find numpy, even though it is already installed in your virtual environment.

This page covers the error raised during an install. If your own code raises ModuleNotFoundError when you run it, see how to fix ModuleNotFoundError: no module named X.

Problem

You’re trying to install a Python package using:

uv pip install git+https://github.com/example/examplepkg.git
# or
pip install .

Even though numpy is installed in your environment, the install fails with:

ModuleNotFoundError: No module named 'numpy'

Why This Happens

By default, pip (and uv pip) use build isolation when installing packages that declare a build backend in pyproject.toml. This means they build the package in a temporary, isolated environment with only the dependencies listed in the [build-system] table of pyproject.toml.

If numpy is required during the build (e.g. to compile C extensions) but not listed in build-system.requires, it will not be available, leading to the error.

Solution

Use the --no-build-isolation flag to bypass the isolated environment and allow access to already-installed packages:

pip install --no-build-isolation .
# or
uv pip install --no-build-isolation git+https://github.com/example/examplepkg.git

This works because --no-build-isolation tells pip to use your current environment, where numpy is already installed. This avoids the problem entirely, but at the cost of potentially inconsistent builds.

Permanent Fix: Add Build Requirements

A better long-term solution is to ensure that required build dependencies like numpy are declared in pyproject.toml, following PEP 518:

[build-system]
requires = ["setuptools", "wheel", "numpy"]
build-backend = "setuptools.build_meta"

This allows pip to correctly build the package in an isolated environment with the necessary dependencies.

Missing build requirements are a common source of install failures. Always specify required build-time packages explicitly in [build-system.requires] to ensure compatibility with modern Python packaging tools.