Turn Your Instagram Saves Into a Content Idea Machine

The pipeline is three pieces:

The full flow: launchd fires sync.py at 9 AM and 9 PM. New saves land in Notion with a "New" status. Once a week, you open Claude Code, type /instagram-sync ideate, approve the ideas you like, and they land in your "Content Ideas" database with hooks, an outline, and platform breakdowns.

One honest point before we build: Instagram exposes no official API for your saved posts, not even through the Graph API. So the script replays the same calls your browser makes, on your own account, with your own session cookies, twice a day. That's low-key and quiet, but automated collection still goes against Meta's terms of use: the risk is low at this frequency, not zero. You build with your eyes open.

Name (title)          Status (select: New, Reviewed, Used)
URL (URL)             Type (select: Post, Reel, Carousel)
Author (text)         Collection (select)
Caption (text)        Media ID (text)
Saved (date)

Name (title)          Pillar (select: your content pillars)
Platform (multi-select: Instagram, TikTok, YouTube)
Format (select: Carousel, Reel, Short, Long-form)
Hook Options (text)   Priority (select: High, Medium, Low)
Angle (text)          Week Of (date)
Status (select: Not started, In progress, Done)

The script authenticates with the session cookies your browser already sends with every request.

Build a Python script `sync.py` that pulls my saved Instagram posts
and writes new ones to a Notion database.

## Inputs (from config.json in the same folder)
ig_session_id, ig_csrftoken, ig_user_id (Instagram web cookies),
notion_token (ntn_ token), notion_database_id (Instagram Saves DB),
collections_filter (optional list of collections to sync).

## Instagram calls (mimic the web client, not the mobile app)
Headers: desktop Chrome User-Agent, X-IG-App-ID: 936619743392459.
Cookies: sessionid, csrftoken, ds_user_id.
Validate the session: GET /api/v1/accounts/edit/web_form_data/
Fetch saves: GET /api/v1/feed/saved/posts/?count=50,
paginate with next_max_id, sleep 1 second between pages.
List collections: GET /api/v1/collections/list/

## Deduplication
state.json holds the media IDs already synced. Skip any post
already present. Only add an ID after a successful Notion write.

## Notion writes (official notion-client SDK)
One page per new post: Name (@author/code), URL, Type (select
based on media_type), Author, Status = "New", Media ID, Saved
(today's date), Caption (truncated to 1900 chars), Collection
when the filter is active.
URL: https://instagram.com/reel/CODE/ for reels, /p/CODE/ otherwise.

## Logs and errors
Log to stdout and sync.log. Final line:
"Sync complete: N new | N skipped | N total | N errors".
Error on one post: log it and keep looping, never crash the run.
A failed ID does NOT go into state.json (retried next run).
Invalid session: exit immediately with a clear message.

## Also generate
requirements.txt (requests, notion-client>=3.0),
config.example.json (placeholders), .gitignore (config.json,
state.json, sync.log, .venv/).

launchd is the Mac's native scheduler: it starts at boot and restarts your job after a reboot, with nothing to maintain.

Generate a launchd file `com.myname.instagram-saves-sync.plist` that
runs `.venv/bin/python3 sync.py` from my project folder at 9:00 AM
and 9:00 PM every day. Absolute paths throughout. stdout to
launchd-stdout.log and stderr to launchd-stderr.log in the project
folder. My project folder: [YOUR_ABSOLUTE_PATH]

Then give me the two commands to install it:
copy to ~/Library/LaunchAgents/ then load with launchctl.

The command is the creative layer of the system. It lives in a markdown file that Claude Code loads as a slash command.

Create a slash command `.claude/commands/instagram-sync.md` that
handles six actions depending on what I ask:

1. Sync now: run `.venv/bin/python3 sync.py`, then move on to
ideation if any saves have Status = "New".

2. Ideation (the core):
- Read the Instagram Saves database (ID: [SAVES_DB_ID]) filtered
  on Status = "New". Past 10 saves, process in batches of 10.
- For each save, generate 1 idea: reframe the concept for my
  audience ([YOUR_NICHE]), assign a pillar based on the source
  collection (my mapping: [COLLECTION to PILLAR]), write 3 hooks
  (curiosity, value, emotion), an outline HOOK then 3-4 key points
  then CTA, and the breakdowns: Instagram (carousel or reel),
  TikTok (30-60s script), YouTube only if the topic has long-form
  depth. Priority: High for core collections or timely topics,
  Medium for strong audience overlap, Low for inspiration.
- Present all ideas and ask: approve all, pick by number, skip
  all, or modify.
- For each approved idea, create a page in Content Ideas
  (ID: [IDEAS_DB_ID]) with every property filled and the full
  outline in the page body.
- Flip each processed save to "Used" (approved) or "Reviewed"
  (skipped).
- End with a summary: saves processed, ideas generated, ideas
  saved, titles.

3. Sync status: read state.json and the last 20 lines of sync.log,
report the last run and recent errors.
4. Scheduler status: `launchctl list | grep instagram-saves`.
5. Refresh session: walk me through grabbing fresh cookies in
DevTools then updating config.json and testing.
6. View recent saves: read the database sorted by date, newest first.

Open your "Content Ideas" database: every approved idea is there, with its angle, its outline, and its platform breakdowns, ready to shoot.