How I Use AI to Turn GitHub Activity into a Weekly Report My Team Actually Reads

Every Friday, my manager asks what we shipped this week. I used to spend 20 minutes scrolling through GitHub, copying PR titles, and formatting a Slack message. Now I say one sentence to AI, and it pulls this week’s merged PRs and closed issues, writes a summary in Notion, and emails it to the team.

What You’ll Learn

• How to pull GitHub PRs and issues from a specific time range
• How AI turns raw commit data into a readable summary
• How to automatically save the report to Notion and email it
• The formatting tricks that make people actually read it

What You Need

• Claude (Pro or above)
• OctoDock account with GitHub, Notion, and Gmail connected
• A GitHub repo you have access to

The Full Workflow

Copy this prompt:

Look at the GitHub repo [owner/repo]. Find all PRs merged this week and all issues closed this week. Write a weekly summary with three sections: what we shipped, what we fixed, and what’s still open. Save it as a Notion page under “Weekly Reports.” Then email it to [[email protected]].

AI executes:

1. github.list_pulls — filters by merged date this week
2. github.list_issues — filters by closed date this week
3. Reads PR descriptions and issue titles, groups them
4. notion.create_page — writes the report
5. gmail.send — emails the report to the team

✅ What success looks like: A Notion page with this week’s activity, neatly organized. Your team gets an email with the same content.

⚠️ Common mistake: If your repo has 100+ PRs this week, AI might hit the API response limit and miss some. Add “only include PRs merged after Monday” to be explicit about the date range.

Tricks That Aren’t Obvious

Tell AI what audience the report is for. “Write it for a non-technical manager” produces very different output than “write it for the engineering team.” The first version says “we improved page load speed by 40%.” The second says “refactored the query layer to use connection pooling.” Same data, completely different value.

Group by category, not by PR number. Raw GitHub data is chronological. But your manager wants to know “what features shipped” and “what bugs were fixed,” not “PR #47, PR #48, PR #49.” In the prompt, ask AI to “group by type: features, bug fixes, infrastructure.”

Include “still open” items. A report that only shows what’s done misses context. Add “list the top 5 open issues by priority” and suddenly your report also serves as a planning tool for next week.

Save every week’s report in the same Notion folder. Over time, you get a timeline of project progress. Useful for quarterly reviews, investor updates, or just remembering what you did three months ago.

What I Ran Into

⚠️ GitHub’s API returns PRs sorted by creation date, not merge date. If a PR was created a month ago but merged today, it won’t appear when you filter by “created this week.” AI needs to fetch recent PRs and then filter by merge date on its side. Sometimes it gets this right, sometimes it misses old PRs that were just merged.

⚠️ Private repos need the right OAuth scope. If AI says “repository not found,” it’s usually a permission issue. Make sure your GitHub connection in OctoDock includes access to private repos.

What This Can’t Do

❌ Can’t analyze code quality. The report covers what was merged, not whether it was good. AI reads PR titles and descriptions, not the actual diff.

❌ Can’t attribute work to specific people. GitHub PRs have authors, but if you’re in an org with many contributors, AI might not know who “jdoe42” is. You’d need to provide a mapping.

vs. Existing Tools

Tools like LinearB, Sleuth, and GitHub’s built-in Insights give you dashboards and metrics. They’re great for long-term trends. But they don’t write a human-readable summary that you can email to your manager.

The OctoDock approach is more like having an intern who reads all your GitHub activity and writes a report in your style. Less polished than a dashboard, but more useful for weekly communication.

Tool Setup

1. Sign up at octo-dock.com
2. Connect GitHub, Notion, and Gmail in the Dashboard
3. Copy your MCP URL → Claude.ai → Settings → MCP Connectors

GitHub OAuth will ask which repos to grant access to. Select the repos you want reports on.