Show HN: LinkedIn-ops – Zero-dependency CDP automation built for AI agents

An independent open-source project called linkedin-ops released a zero-dependency LinkedIn automation engine and CLI that drives a local Google Chrome instance over the Chrome DevTools Protocol (CDP) on port 9222, aimed at AI agents and multi-agent workflows. The tool ships an npm-installable setup wizard that locates the Chrome binary and existing profiles, a doctor command that checks CDP reachability and session authentication, and a cross-process mutex lock (linkedin.lock) plus isolated ephemeral tabs with auto-cleanup to serialize concurrent agents and prevent session collisions. Commands cover connection invites, direct messages, post comments and company people search, with dry-run defaults and a --send flag for live submission; the project states it is not affiliated with LinkedIn and warns users to respect platform terms and rate limits.

High-reliability, multi-agent-safe LinkedIn automation engine and CLI built on Chrome DevTools Protocol CDP . Disclaimer & Responsible Use : linkedin-ops is an independent open-source tool built for developer automation research, testing, and workflow automation via local Chrome DevTools Protocol CDP . This project is not affiliated with, endorsed by, or sponsored by LinkedIn Corporation. Automated interaction with LinkedIn must comply with applicable platform terms and policies. Always maintain realistic human delays and respect LinkedIn rate limits to safeguard account health. npm install node index.mjs setup The setup wizard automatically: 1. Locates your Google Chrome binary macOS, Linux, Windows . 2. Discovers your existing Chrome profiles from system User Data so you can pick your logged-in account or create a dedicated project session . 3. Saves settings to .linkedin-opsrc no manual path pasting required . 4. Generates a one-click launcher start-chrome.command / start-chrome.bat . npm run doctor or: node index.mjs doctor Checks CDP port 9222 reachability, browser version, user agent, and verifies an authenticated LinkedIn session. node index.mjs connect --profile="https://www.linkedin.com/in/username" --note="Hi, loved your post on distributed systems..." Dry-run mode navigates and fills the note without submitting. Add --send only when ready to live submit . 1. Automated Runtime Fallback : Daily commands ping port 9222. If Chrome is closed, linkedin-ops automatically launches it using your configured profile. If Chrome is already running without remote debugging enabled, it detects the profile lock and prompts to restart cleanly. 2. Cross-Process Mutex Locking : Serializes execution across concurrent agents and processes linkedin.lock . Prevents session collisions, focus-stealing, and rate-limiting. 3. Isolated Ephemeral Tabs : Each operation spawns its own dedicated Chrome tab via PUT /json/new and connects directly via target ID. 4. Auto-Cleanup : Guarantees tab destruction /json/close/<targetId in a finally block on success or failure, eliminating zombie tab leaks. Automated setup wizard node index.mjs setup Pre-flight health check node index.mjs doctor Connection invitation with personalized note dry-run node index.mjs connect --profile="https://www.linkedin.com/in/username" --note="Hi, loved your post..." Live connection invitation node index.mjs connect --profile="https://www.linkedin.com/in/username" --note="Hi, loved your post..." --send Direct message to 1st-degree connection node index.mjs dm --profile="https://www.linkedin.com/in/username" --message="Checking in on our conversation." --send Comment on a LinkedIn post node index.mjs comment --url="https://www.linkedin.com/posts/..." --text="Great insights on distributed systems " --send Company people search e.g. recruiters node index.mjs search --company="Google" --keywords="recruiter" --limit=10 --json Inspect or force-release active mutex lock node index.mjs lock node index.mjs lock release js import { withLock, withDedicatedTab, sendConnectionInvite } from 'linkedin-ops'; await withLock 'outreach-task', async = { await withDedicatedTab { port: 9222 }, async cdp = { const result = await sendConnectionInvite cdp, { linkedinUrl: 'https://www.linkedin.com/in/username', note: 'Hi, loved your work on distributed systems.', send: true } ; console.log result ; } ; } ; For multi-profile outreach and candidate sweeps, see campaigns/batch-connect.example.mjs https://github.com/shashtag-ventures/linkedin-ops/blob/main/campaigns/batch-connect.example.mjs : Copy template to your personal user campaigns folder cp campaigns/batch-connect.example.mjs campaigns/my-outreach.mjs Run the campaign safely with mutex locking and isolated tab lifecycles node campaigns/my-outreach.mjs User campaigns and harvested datasets in data/ and campaigns/ are gitignored and never committed to source control . - DATA CONTRACT.md https://github.com/shashtag-ventures/linkedin-ops/blob/main/DATA CONTRACT.md : Strict definition of the System Layer vs User Layer boundaries. - Architecture Blueprint https://github.com/shashtag-ventures/linkedin-ops/blob/main/docs/ARCHITECTURE.md : Deep dive into CDP socket lifecycle, locking mechanisms, and tab management.