How to Run an AI GitHub Productivity Agent with OE Runtime

Get a clear picture of your GitHub repository's health in seconds. OE Runtime fetches open issues and pull requests, triages by age and label, flags what is stale or unreviewed, and delivers a prioritised action list — no manual triage required.

The Agent YAML

Create a file called agent.yaml inside a productivity-crm/ directory:

name: GitHub Assistant
description: Summarize open issues and PRs in a repository
instructions: |
  You are a GitHub assistant. Fetch repository activity, summarize open issues and pull requests,
  and identify what needs immediate attention.
  Complete all steps fully before writing your report.
steps:
  - name: Fetch Repository Activity
    content: |
      Using the GitHub connector, fetch:
      - All open issues (up to 20), sorted by most recently updated
      - All open pull requests (up to 10), sorted by most recently updated
      For each, note: title, author, labels, created date, and days open.
  - name: Triage and Prioritize
    content: |
      From the fetched issues and PRs:
      - Flag any issue or PR open for more than 14 days as stale
      - Group issues by label (bug, enhancement, question, other)
      - Identify PRs that have no reviewers assigned
      - Find any issues marked as "bug" that have no assigned PR
  - name: Report
    content: |
      Produce a repository health summary:
      - Open issues: total, breakdown by label, stale count
      - Open PRs: total, unreviewed count, stale count
      - Top 3 most urgent issues (bugs with no PR)
      - Recommended next actions
connectors:
  - connection_name: GitHub
    connection_type: github

The Config File

Create oe-config.json in the same directory:

{
  "llm": {
    "provider": "openai",
    "model": "gpt-4o",
    "apiKey": "YOUR_OPENAI_API_KEY"
  },
  "server": {
    "enabled": false,
    "port": 3333,
    "apiKey": "your-secret-api-key"
  },
  "connectors": [
    {
      "connection_name": "GitHub",
      "connection_type": "github",
      "token": "ghp_YOUR_GITHUB_PERSONAL_ACCESS_TOKEN",
      "owner": "your-org",
      "repo": "your-repo"
    }
  ]
}

Create a Personal Access Token at github.com/settings/tokens with repo scope. Replace your-org and your-repo with your organisation and repository names.

Download OE Runtime

Run the Agent

Windows

# Run the agent
oe-runtime-win.exe productivity-crm/agent.yaml --config productivity-crm/oe-config.json

macOS

# Make executable (first time only)
chmod +x oe-runtime-macos

# Run the agent
oe-runtime-macos productivity-crm/agent.yaml --config productivity-crm/oe-config.json

Linux

# Make executable (first time only)
chmod +x oe-runtime-linux

# Run the agent
oe-runtime-linux productivity-crm/agent.yaml --config productivity-crm/oe-config.json

Running as an API Server

Add --serve to start OE Runtime as an HTTP server on port 3333:

oe-runtime-win.exe --serve --config productivity-crm/oe-config.json

Then send a request:

curl -X POST http://localhost:3333/run \
  -H "Content-Type: application/json" \
  -d '{"yaml": "productivity-crm/agent.yaml", "params": {}}'

Or import the Postman collection (download above) to test all endpoints visually.

Use Cases

• Daily repository health checks — Run the agent each morning to get a snapshot of open issues, PR status, and stale items across your repository. The report tells your team what to focus on without anyone needing to manually scan the issue tracker or pull request queue.
• Sprint planning support — Before a sprint planning session, run the agent to produce a triaged list of open bugs, feature requests, and technical debt items with their age and labels. This gives your team a prioritised backlog view without the overhead of manual issue grooming.
• Pull request review management — Identify pull requests that have been open for more than a week with no reviewer assigned and surface them as a priority. Unreviewed PRs slow down the team and block feature delivery — the agent flags them so engineering leads can reassign or expedite review.
• Release readiness assessment — Before cutting a release, run the agent to confirm there are no open bug issues without an associated pull request and that all critical PRs have been reviewed and merged. This gives your release manager confidence that nothing critical has been missed.
• Open source community management — For open source maintainers, the agent provides a regular digest of unanswered community questions, stale contributor PRs, and unfiled bug reports. It helps maintainers stay on top of community engagement even with limited time.
• Cross-repository portfolio view — Run the agent against multiple repositories in sequence to build a portfolio-wide health report for engineering leadership. Each report shows the issue and PR status per repo, giving a consolidated view of development health across the entire product.
• Automated triage notifications — Run the agent as an API server and integrate it with your Slack or email setup to deliver automated triage summaries to the relevant channel or inbox. Engineering managers get a daily digest without any manual reporting effort.
• Onboarding for new contributors — Point new developers at the agent to get an instant view of what is most urgently needed in the repository. The prioritised issue list with bug tags and PR gaps gives newcomers a clear starting point for their first meaningful contribution.