LinkedIn Feed Tracker — Cowork Plugin

See what your network is really talking about.

Collect every post from your LinkedIn connections. Claude does the rest: finds trends, spots content gaps, tracks your own post performance, and creates weekly intelligence reports. One setup file. Then just ask questions.

01
How It Works
Collect, store, analyze, repeat
02
Setup
One file. Double-click. Done.
03
What You Get
Dashboards, reports, deep dives
04
Talk to Claude
Examples of what to ask
05
Privacy
Your data, your machine
06
FAQ
Questions answered
1
How It Works

Four collectors, one database, Claude on top

LinkedIn Feed Tracker is a Cowork hybrid plugin. Automated scrapers collect data from your LinkedIn every night. Claude analyzes everything inside Cowork. All data stays on your Mac.

The collection layer

Four scrapers run on your Mac (via a nightly schedule) and capture different slices of your LinkedIn world:

Feed Collector

Scrolls your LinkedIn home feed and captures every post—author, content, engagement counts, timestamps. Deduplicates automatically.

Runs nightly

Profile Collector

Visits your connections' profiles on a 7-day rotation. Each night it covers ~1/7 of your network, so every connection gets visited once a week.

Rotates nightly

Connections Sync

Keeps your full connections list up to date. Handles networks up to 5,000+ with a multi-strategy approach that auto-adapts to LinkedIn changes.

Runs with setup + on demand

Own Post Tracker

Scrapes your own posts' performance—likes, comments, reposts—so you can see what's working and track growth over time.

Runs nightly

The intelligence layer

💾

Local SQLite Database

All posts, connections, and performance data live in a single database file in your workspace. You choose where it's stored during setup.

🤖

Claude Analyzes

Open Cowork and ask anything. Claude queries your database, finds trends, generates dashboards, and writes reports—all from your own data.

Self-healing selectors

LinkedIn changes their page layout frequently. The plugin detects when its selectors break, dumps diagnostic info, and flags itself for repair. Next time you open Cowork, Claude auto-fixes the selectors. You don't have to do anything.

Why this is powerful

You're not relying on LinkedIn's algorithm to tell you what matters. You're asking your data what matters, with Claude as your analyst.

2
Setup

One file. Double-click. Everything happens.

Step 0: Install the plugin

Before anything else, you need to add the plugin to Cowork:

1.
Open Claude Desktop and go to Settings → Plugins.
2.
Search for LinkedIn Feed Tracker and click Install. This adds the plugin files to your Cowork workspace.
3.
Open the workspace folder the plugin was installed into. You'll see a Setup folder with the .command files referenced below.
Node.js 20 is handled automatically

The setup files install and configure Node.js 20 via nvm if it's not already on your machine. You don't need to install it manually. A key dependency (better-sqlite3) requires Node 20 specifically—not 22—on Apple Silicon Macs.

Setup (4 double-clicks)

Open the Setup folder inside your LinkedIn Feed Tracker workspace. You'll see four .command files. Double-click them in order:

1
Login to LinkedIn.command
Opens a browser window. Log in to LinkedIn normally. The tool saves your session locally so future runs are headless. If LinkedIn asks for a verification code, complete it in the browser—same as normal.
2
Sync Connections.command
Downloads your full LinkedIn connections list. This powers the profile-scraping feature—it rotates through 1/7 of your connections every night so everyone gets visited weekly.
3
Collect Now.command
Runs your first feed collection. Scrolls your LinkedIn feed, parses every post, and saves them to the local database. Takes 1–2 minutes.
4
Install Nightly Automation.command
This is the key step. It copies the collector to your local disk (~/.linkedin-feed-tracker/), installs a macOS LaunchAgent, and activates nightly collection at 10 PM. Takes 30 seconds.
If a .command file won't open

macOS blocks unsigned scripts. Right-click the file and choose "Open" instead of double-clicking. If that doesn't work, go to System Settings → Privacy & Security and click "Open Anyway."

Why Step 4 matters

Node.js + SQLite can't run reliably on iCloud Drive—file locking and sync cause crashes. Step 4 solves this by copying the collector to your local disk (~/.linkedin-feed-tracker/) and running from there. After each nightly collection, the script copies the database back to iCloud so Cowork can read it.

How the nightly pipeline works

Every night at 10 PM, macOS runs nightly.sh which: (1) syncs code from iCloud to local disk, (2) runs the feed collector, (3) runs the profile batch scraper, (4) checkpoints the SQLite WAL, and (5) copies feeds.db back to iCloud. Logs go to ~/.linkedin-feed-tracker/logs/ and auto-delete after 14 days.

That's it

After setup, data flows automatically every night. Open Cowork anytime and start asking Claude about your network.

After setup

Your data lives in two places: ~/.linkedin-feed-tracker/app/data/feeds.db (primary, on local disk) and a synced copy on iCloud that Cowork reads. You can trigger a manual collection anytime by double-clicking Collect Now.command or running bash ~/.linkedin-feed-tracker/nightly.sh in Terminal.

Set your topic clusters

The plugin ships with default topics, but you'll want to customize them. Open Cowork and tell Claude: "Show my topics" to see what's configured, then "Add a topic for [whatever you care about]" or "Remove the topic about [X]." Claude handles it instantly. This is how you focus the analysis on what matters to you.

3
What You Get

Dashboard, reports, analysis, and performance tracking

Once data is flowing, Claude can answer any question you ask inside Cowork. Here's what the plugin gives you:

Interactive Dashboard

Ask Claude: "Show me my dashboard"

Claude builds a live HTML dashboard from your database. Here's what it looks like:

Connections
1,966
Synced
Feed Posts
847
Collected
Your Posts
23
Tracked
Days Active
14
Tracking
Your Post Performance
Post Types
Original 68%
Image 22%
Repost 7%
Article 3%
Example dashboard — yours is generated live from your data

Weekly Intelligence Report

Ask Claude: "Generate a weekly report"

A formatted Word document that includes:

  • What topics trended in your network this week
  • Which connections drove the most engagement
  • Emerging patterns and rising topics
  • Content gaps nobody in your network is covering
  • Recommended post ideas based on gaps and trends

Feed Analysis

Ask Claude: "Analyze my feed"

Deep analysis of collected posts: topic clustering, engagement breakdowns, author activity, and trend spotting. Goes beyond what the dashboard shows with narrative insights.

Own Post Performance

Ask Claude: "How are my posts doing?"

Since the plugin tracks your own post performance (likes, comments, reposts), Claude can tell you what's working, compare against your baseline, and identify which topics and formats get the best engagement from your specific audience.

Your Posts
42
Tracked
Avg. Likes
38
+12% vs baseline
Avg. Comments
7
+22% vs baseline
Engagement Over Time
4 wks ago2 wks agoNow
Likes Baseline
Best Performing
1 WuunderFund CEO announce
45 comments · 9 reposts
2 Bay Area cloud + AI event
14 likes · 0 comments
3 Hiring — join the team
11 likes · 0 comments
Example own-post metrics — yours are generated live from your data

Topic Clusters

Your network's posts are auto-sorted into configurable topic clusters. Manage them by telling Claude: "Add a topic" or "Show my topics."

Work Topics

Remote work, careers, hiring, skill development, AI at work

Industry Topics

HR & benefits, payroll & fintech, employee retention, travel & lifestyle benefits

Startup Talk

Fundraising, founder advice, product, SaaS, growth

Personal

Life updates, celebrations, reflections, wellness

4
Talk to Claude

Ask questions like you're texting a friend

Once data is flowing, open Cowork and ask Claude anything about your network. You don't need special commands. Just talk naturally.

Example conversations

What's trending in my network right now?
Based on this week's posts, the top topics are: remote work culture (23 posts), AI tooling (18 posts), and hiring strategies (15 posts). Remote work is up 40% from last week. I can show you the dashboard or dive into any of these.
Show my dashboard.
[Generates an interactive HTML dashboard with charts, filters, and top-level metrics]
How did my last post perform?
Your post about hiring challenges got 42 likes and 11 comments—that's 35% above your 30-day average. Comments were mostly from HR and recruiting connections. Your question-style hook worked well.
What's trending in HR and Employee Benefits?
Top posts in that cluster are about flexible work arrangements, mental health support, and equity compensation. Three of your connections in HR posted about benefits that improve retention—there might be an angle for you here.
Generate my weekly report.
[Builds a formatted Word document with this week's trends, top posts, engagement patterns, and content recommendations]

You can also use slash commands

  • /show-dashboard — Generate and display your analytics dashboard
  • /collect-now — Trigger a manual data collection
  • /health-check — Check system health (database, collection status, errors)

Other things to ask

  • "Who's been most active this month?"
  • "What topics did I post about recently?"
  • "Who posted about employee retention?"
  • "What's the engagement breakdown by topic?"
  • "Show me posts about payroll and fintech."
  • "Add a topic cluster for AI in recruiting."
5
Privacy

Your data stays on your machine

What gets collected

  • Post text and metadata (timestamps, engagement counts)
  • Author names and connection info
  • Your own post performance metrics
  • Your LinkedIn browser session (stored locally for automated login)

Where it's stored

Everything lives on your Mac. The scrapers run locally. The database is a single file in a folder you choose. There's no cloud component, no external server, no data transmission.

Who sees it

  • You
  • Claude (only when you ask questions inside Cowork)
  • No one else

What about LinkedIn?

The scraper uses a real browser (Playwright with Chromium) and mimics normal browsing. LinkedIn sees standard page visits, not API calls or bulk requests. Your credentials stay in your local browser profile—they're never transmitted anywhere.

You're in control

Delete the database anytime. Pause the nightly schedule anytime. The plugin is open-source—you can review every line of code on GitHub.

6
FAQ

Questions answered

Do I need to know how to code?

No. You double-click one setup file and then talk to Claude. That's it.

What platforms are supported?

macOS only right now. The scrapers use macOS .command files and the nightly schedule uses macOS launchd. Windows and Linux support may come later.

What if LinkedIn changes their website?

The plugin has a self-healing selector system. When LinkedIn changes their DOM, the scraper detects the failure, dumps diagnostic info, and flags itself for repair. Next time you open Cowork, Claude fixes the selectors automatically. You don't do anything.

How often does it collect data?

Every night at 10 PM automatically via a macOS LaunchAgent. The nightly run syncs code to local disk, collects your feed, visits a batch of connection profiles, and copies the database back to iCloud. If your Mac is asleep at 10 PM, macOS catches up when it wakes.

What if I have thousands of connections?

The connection sync handles networks up to 5,000+ using a 4-strategy fallback system. For ongoing monitoring, the profile collector uses a 7-day rotation—it visits ~1/7 of your connections each night so everyone gets covered weekly.

Can I trigger a collection manually?

Yes. Either double-click Collect Now.command in the Setup folder, run bash ~/.linkedin-feed-tracker/nightly.sh in Terminal, or tell Claude /collect-now inside Cowork.

Can I export my data?

Yes. The database is a standard SQLite file—you can query it with any SQLite tool. You can also ask Claude to generate reports as Word documents, CSVs, or dashboards you can share.

Will this slow down my computer?

The scrapers run briefly (a few minutes) and then stop. Outside of collection windows, there's zero background load. The nightly schedule only triggers once at 10 PM.

Why does the collector run on local disk instead of iCloud?

Node.js + SQLite can't run reliably on iCloud Drive—file locking and sync cause EAGAIN errors that crash the collector. Step 4 of setup solves this by copying the project to ~/.linkedin-feed-tracker/ (local disk), running the collector there, and copying feeds.db back to iCloud after each run so Cowork can read it.

Where is my data stored?

The primary database lives at ~/.linkedin-feed-tracker/app/data/feeds.db on your local disk. After each nightly run, it's copied to two iCloud locations: your LinkedIn Feed Tracker workspace folder and the CLAUDE OUTPUTS backup. Cowork reads from iCloud. Your browser session (LinkedIn cookies) is at ~/.linkedin-feed-tracker/browser-profile/.

How do I check if the nightly collection is working?

Check today's log: open Terminal and run cat ~/.linkedin-feed-tracker/logs/nightly-$(date +%Y-%m-%d).log. Or run launchctl list | grep com.lft to confirm the LaunchAgent is loaded. You can also ask Claude /health-check inside Cowork.

How do I change the collection time?

Edit ~/Library/LaunchAgents/com.lft.nightly.plist—change the Hour (0–23) and Minute (0–59) values under StartCalendarInterval. Then reload: launchctl unload ~/Library/LaunchAgents/com.lft.nightly.plist && launchctl load ~/Library/LaunchAgents/com.lft.nightly.plist

Why Node.js 20 and not 22?

A key dependency (better-sqlite3) doesn't have a prebuilt binary for Node.js 22 on Apple Silicon Macs. The setup file handles this automatically—it installs Node 20 via nvm and sets it as your default. You don't need to manage versions yourself.

Does LinkedIn know I'm scraping?

LinkedIn sees normal browser traffic. The plugin uses a real Chromium browser with your actual session—it's visiting pages the same way you would manually. There are no API calls or unusual request patterns.

How much does this cost?

The plugin is free and open-source. You need a Cowork subscription for the Claude-powered analysis features, but that's your regular Anthropic subscription.

What if I get logged out of LinkedIn?

Double-click Login to LinkedIn.command in the Setup folder to re-authenticate. Takes about 30 seconds. Sessions typically last 1–2 weeks.

What if the .command file won't open?

macOS blocks unsigned scripts by default. Right-click the file, select "Open," then click "Open" in the confirmation dialog. You only have to do this once per file.

Is this part of a bigger system?

Yes. LinkedIn Feed Tracker is Plugin 1 of 3. Two more plugins are in development that build on top of the data this one collects. More details coming soon.

Still stuck?

Check the GitHub repo for issues and updates, or reach out to the creator.

Coming Soon

This is Plugin 1 of 3.

Two more plugins are in development that build on top of your data—turning network intelligence into content and predictions. More details soon.

Ready to see what your network is talking about?

Setup takes about 10 minutes. Data starts flowing that night.