Run GitHub Actions workflows locally in Docker to test jobs, event payloads, and CI changes before pushing.
$brew install act
Agent Compatibility
JSON Output
Agent Skill
MCP Support
AI Analysis
act is a local runner for GitHub Actions workflows. It reads .github/workflows, plans jobs for an event or selected job, and executes them in Docker so you can test CI behavior before pushing.
What It Enables
- Run workflow jobs locally against a chosen event payload or job ID, with local secrets, vars, env files, matrix filters, and repository overrides.
- List workflows, inspect stage ordering as a graph, validate workflow definitions, and dry-run execution paths before spending CI time on GitHub-hosted runners.
- Iterate faster on workflows and custom actions with local caches, artifact or cache servers, and
--watchreruns when files change.
Agent Fit
- Once Docker access and runner image mappings are configured, commands are mostly non-interactive and fit inspect-change-verify loops around CI debugging.
- Structured output is limited but real:
--jsonemits JSON logs and--list-optionsreturns a JSON description of supported flags. - Best for agents already operating inside a repo with GitHub Actions; it is less useful as a general GitHub control plane because behavior depends on local workflow fidelity and containerized runner support.
Caveats
- Execution depends on Docker-compatible containers and suitable runner images, and the docs explicitly note that
actis not completely compatible with GitHub runners. - First run can prompt for a default image selection unless
.actrcor explicit-Por--platformmappings are already in place.