If your Python projects pin dependencies with pip-compile and pip-sync, the latest release deserves a few minutes of your time. pip-tools 7.7.0, released on 15 November 2026, moves the project from Jazzband to the Python Packaging Authority (PyPA), raises the minimum pip version to 26.2.2, adds a new stable extra, and fixes a bug that made dependency resolution slower than it should be.

This guide covers what changed, how to upgrade safely in local development and CI, and how to use the new stable extra when a fresh pip release breaks your build. Sources: the official changelog, the release post by maintainer Stephen Rosen and the PyPI metadata.

Quick refresher: what pip-tools does

pip-tools keeps two files apart:

  • requirements.in (or pyproject.toml) lists the packages you actually want.
  • requirements.txt is the fully pinned, generated lock file.
# requirements.in
fastapi
httpx>=0.27
pip-compile requirements.in          # writes requirements.txt with exact pins
pip-sync requirements.txt            # makes the venv match the lock file exactly

pip-sync installs what is missing and removes what is not listed, so every developer and every CI job ends up with the same environment.

What’s new in pip-tools 7.7.0

1. The project now lives under PyPA

The repository moved to github.com/pypa/pip-tools and the docs moved to pip-tools.pypa.io. The maintainers are the same people. A few administrative things changed:

  • The project now follows the PSF code of conduct.
  • Security reports go through a new policy and GitHub Security Advisories, not Jazzband.
  • Sponsorship is directed to PyPA or to the maintainers directly.

Action for you: update any bookmarks, Renovate/Dependabot rules or internal docs that point at the old jazzband/pip-tools URL.

2. pip 26.2.2 or newer is now required

This is the change most likely to affect you. pip-tools 7.7.0 declares pip>=26.2.2 as a dependency, and support for older pip versions has been removed.

The release post explains why: older pip versions bundle root certificates that are now considered unsafe, and a higher floor makes testing and maintenance simpler. The maintainers are careful to say the bump alone does not guarantee a secure install, because whatever tool installs the upgrade must itself be trustworthy. But once you have installed it securely, you will not be left on a known-insecure pip.

3. Python 3.9 dropped, Python 3.15 added

pip-tools 7.7.0 requires Python 3.10 or newer (requires_python >=3.10 on PyPI) and adds Python 3.15 support. If you still have a Python 3.9 CI job, it will stay on an older pip-tools release.

4. A new stable extra

python -m pip install "pip-tools[stable]"

The stable extra caps pip at the latest version tested in pip-tools CI at release time. For 7.7.0 the PyPI metadata shows that cap as pip<=26.3. Most users don’t need it, but it is a handy escape hatch when a new pip release breaks pip-compile.

5. Faster resolution again

Version 7.6.1 introduced a bug in how pip-tools built PyPI JSON API URLs (it dropped /pypi/ from the path). As a result, metadata lookups fell back to downloading packages, which used more network and disk. 7.7.0 fixes this, so on PyPI and other indexes that offer the JSON API you get the fast path back.

6. Smaller fixes

  • pip-sync no longer uninstalls pip when pip was installed from a direct URL.
  • The docs now explain how to control “unsafe” packages with --allow-unsafe and --unsafe-package.

Upgrading locally

Upgrade pip first, then pip-tools, inside each virtual environment:

python -m pip install --upgrade "pip>=26.2.2"
python -m pip install --upgrade pip-tools
pip-compile --version

Then regenerate your lock file and check the diff:

pip-compile --upgrade-package pip-tools requirements-dev.in
git diff requirements-dev.txt

If you only changed the tooling, your application pins should not change. If they do, review them before committing.

Upgrading in CI

Most CI breakage comes from old pip versions baked into images. Make the order explicit:

# .github/workflows/lock-check.yml
name: lock-check
on: [pull_request]

jobs:
  check:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-python@v5
        with:
          python-version: "3.13"
      - name: Install tooling
        run: |
          python -m pip install --upgrade "pip>=26.2.2"
          python -m pip install "pip-tools==7.7.0"
      - name: Verify lock file is up to date
        run: |
          pip-compile --quiet --output-file=/tmp/requirements.txt requirements.in
          diff -u requirements.txt /tmp/requirements.txt

Pinning pip-tools==7.7.0 in CI is a good habit: tooling upgrades then happen in a reviewed pull request, not by surprise.

Using the stable extra as a safety valve

Suppose a new pip release lands and your pip-compile step starts failing. Instead of hunting for the last working pip version by hand, switch the install line:

python -m pip install "pip-tools[stable]==7.7.0"

This pulls in a pip version no newer than the one pip-tools tested. Remove the extra again once a pip-tools release supports the new pip.

Compiling from pyproject.toml

If your dependencies live in pyproject.toml, pip-tools can read them directly:

pip-compile pyproject.toml --output-file requirements.txt
pip-compile pyproject.toml --extra dev --output-file requirements-dev.txt

This works the same in 7.7.0; just make sure the environment running pip-compile meets the new Python and pip minimums.

Upgrade checklist

  1. Confirm every environment that runs pip-tools uses Python 3.10+.
  2. Upgrade pip to 26.2.2 or newer before installing pip-tools 7.7.0.
  3. Update links from jazzband/pip-tools to pypa/pip-tools.
  4. Regenerate lock files and review the diff.
  5. Pin the pip-tools version in CI and keep pip-tools[stable] in mind for emergencies.
  6. If you skipped 7.6.1’s slow lookups, enjoy faster pip-compile runs again.

Takeaway

pip-tools 7.7.0 is mostly a housekeeping release, but the new pip floor can break older images and Python 3.9 jobs. Upgrade pip first, pin pip-tools in CI, and use the stable extra if a future pip release causes trouble. The full list of changes is in the pip-tools changelog.