Generating code

dagger generate

Runs every generator in your workspace. A generator doesn't write to your files directly — it returns a changeset: a diff of the proposed changes. Dagger shows you the changed paths and line counts, and nothing is written until you approve.

Functions can also return a changeset without being generators — a formatter's fix, called with dagger api call, is the common case. Checks never do; they only validate.

List generators

dagger generate -l

Filter generators

dagger generate protobuf:*         # all generators from a module
dagger generate changelog:generate  # a single generator

Apply without prompting

Pass -y / --auto-apply to skip the review step — useful in scripts and non-interactive sessions:

dagger generate -y

Coding agents

When Dagger detects that a coding agent is running dagger generate, it requires an explicit choice up front. Pass -y to apply the result, or --no-apply to run the generators and show the changes without writing them:

dagger generate --no-apply

--no-apply exits successfully even when there are pending changes, just like choosing Discard at the interactive prompt. Generators still run and may perform other work; only the changeset is withheld.

Verify in CI

In CI you usually want to check that committed files are up to date rather than rewrite them. dagger check runs each generator as a read-only check and fails, without applying anything, if its output differs from what's committed:

# GitHub Actions
- run: dagger check --generate

A failing generator check means the committed output is stale — run dagger generate locally, apply the changeset, and commit.