How to Speed Up Checkout of a Large Repository in CI

The biggest checkout wins are fetch-depth: 1, filter: blob:none, a scoped sparse-checkout, and turning off tag fetching.

Checkout time on a large repo is a stack of independent levers. Combine a shallow depth, a partial clone filter, a sparse working tree, and disabled tag fetching to get the smallest, fastest checkout a job can run with.

Checkout tuning table

Input	Effect	Use when
`fetch-depth: 1`	clone only the tip commit	no history needed (default)
`filter: blob:none`	defer blob download	need history but not all files
`sparse-checkout`	materialize only some dirs	job builds one area
`fetch-tags: false`	skip fetching tags	many tags, none needed

Combined workflow

.github/workflows/ci.yml

steps:
  - uses: actions/checkout@v4
    with:
      fetch-depth: 1
      filter: blob:none
      fetch-tags: false
      sparse-checkout: |
        services/api

Gotchas

Do not stack flags a downstream step contradicts; a release job that reads tags cannot use fetch-tags: false.
Latchkey managed runners keep a warm git cache and auto-retry transient fetch failures, so cold clones hurt less.

How to Speed Up Checkout of a Large Repository in CI

Checkout tuning table

Combined workflow

Gotchas

Related guides