Never miss a beat. Automatically surface and summarize what's actually happening in your codebase, powered by GitHub Agentic Workflows.
$ gh aw add-wizard e-straight/heartbeat/reporting e-straight/heartbeat/daily-summary
Get started
Install in under five minutes. No code changes, no app installations, no third-party services.
Use the gh-aw extension to add the workflow to your repository.
gh aw add-wizard e-straight/heartbeat/reporting e-straight/heartbeat/daily-summary
You need a personal access token so the workflow can post to your project board. Click the button below to create one with the correct scopes (project and read:org), copy it, then paste it into the command.
Create token (scopes pre-selected)# Set the token you just created
gh aw secrets set GH_AW_PROJECT_GITHUB_TOKEN --value "ghp_your_token_here"
Pick the GitHub project board you want the heartbeat posted to.
# Select your target project
data=$(gh api graphql \
-f query='{ viewer { projectsV2(first: 20) { nodes { title number url } } } }' \
--jq '.data.viewer.projectsV2.nodes[] | "\(.title)\t\(.url)"')
IFS=$'\n'
select label in $(echo "$data" | cut -f1); do
url=$(echo "$data" | sed -n "${REPLY}p" | cut -f2)
repo=$(gh repo view --json nameWithOwner -q .nameWithOwner)
gh variable set GH_AW_TARGET_PROJECT -b "$url" -R "$repo"
break
done
Workflows
Install one or both to keep your GitHub Project board in sync with everything happening in your codebase.
Every meaningful repository event triggers an AI-generated status update on your GitHub Project board.
A nightly digest of all status updates from the past 24 hours, grouped by category.
Before you begin
Make sure you have these tools installed before running the setup commands.
gh), version 2.0 or latergh-aw extension (install below)gh extension install github/gh-aw
Why agentic workflows
The gh-aw framework treats AI agents as untrusted by default and enforces security at every layer.
Agents run with read-only permissions by default. Write operations are never granted directly. They flow through safe-output jobs with scoped, validated access.
Agent output is captured as structured data, scanned for threats like secret leaks and injection attacks, then executed by separate jobs, not by the agent itself.
Markdown workflows compile to pinned .lock.yml files with SHA-locked actions, validated schemas, and hardened configs. What you review is what runs.