How to Migrate Self-Hosted Agents to Self-Hosted Runners

A Jenkins agent or GitLab runner becomes a GitHub Actions self-hosted runner; the agent label or runner tag becomes a runner label you target with runs-on.

Register a self-hosted runner on the repo or organization, label it to match the capability you relied on (for example gpu or arm64), then target it with runs-on: [self-hosted, <label>].

Concept mapping

Other systems	GitHub Actions
Jenkins agent + `label`	self-hosted runner + label
GitLab runner + `tags`	self-hosted runner + label
`agent { label 'gpu' }`	`runs-on: [self-hosted, gpu]`
runner concurrency	runner group + count

Register and target

Terminal

# On the machine, from repo Settings > Actions > Runners
./config.sh --url https://github.com/my-org/my-repo --token <REG_TOKEN> --labels gpu,linux
./run.sh

Workflow

.github/workflows/ci.yml

jobs:
  train:
    runs-on: [self-hosted, gpu]
    steps:
      - uses: actions/checkout@v4
      - run: ./train.sh

What does not map cleanly

Self-hosted runners do not auto-scale; you size and patch the fleet yourself, or use managed runners such as Latchkey that scale and auto-heal transient failures.
Runner labels are advisory; a mislabeled runner can pick up a job it cannot serve.
Ephemeral runs need the --ephemeral flag plus fresh registration each job.