LinkedIn lead generation
Google the hottest LinkedIn posts on "AI agents that browse the web", pull profiles of people who reacted/commented, and flag those matching "CTO building AI agents at a US startup".
Growth hacking
Find 10 job title at kind of company in location on LinkedIn, read each profile, and summarize their current role and their latest post.

LinkedIn is where you check who someone is, what they work on and what they said last week. An AI agent can read your email and your CRM, but when you ask it about a person on LinkedIn, it searches the web and finds a stale snippet, or nothing at all.
LinkedIn gives an agent no way in. It publishes no MCP server, and its self-serve API stops at signing in and posting on your own profile: no people search, no other members' profiles, no inbox. So the LinkedIn MCP servers fall into two groups. The API-based ones (Composio, Zapier, the content tools) publish posts and read your own analytics. The ones that read profiles drive a logged-in LinkedIn session, either on your machine, with a Python or Docker install that only local clients can reach, or on the vendor's servers, with your session held there.
Reduck is a hosted MCP server whose LinkedIn scripts run in your own Chrome, where you are already signed in to LinkedIn. Your agent sees what you see: people search with LinkedIn's filters, full profiles with experience and education, any member's recent posts, post search, your inbox, and messages to your connections. It sends nothing unless you ask. There is no developer app, no install on your machine and no second LinkedIn session somewhere else.
| Script | What it does | Takes |
|---|---|---|
search_people |
People matching keywords and filters, 10 per page, with headline, location, degree and mutual connections | keywords, title, company, location id, page |
get_profile |
A profile's headline, about, location, followers, experience and education | a public id |
get_profile_experience |
Every position on a profile, one entry per role | a public id |
get_profile_posts |
A member's latest posts, with reactions, comments and reposts | a public id, how many |
search_posts |
Posts matching keywords, by relevance or newest first, with their authors and counts | keywords, how many |
get_post_reactors |
Who reacted to a post | its url |
list_inbox |
Your latest conversations, with the last message and the unread count | how many |
get_thread |
The messages of one conversation | its url |
send_message |
A message in a conversation or to a connection, with a dry run | a thread or profile url, the text |
Every script answers structured data: names, profile links, positions, dates, the text as LinkedIn shows it and its counts. The summary and the reasoning stay in your agent. The full list of LinkedIn scripts is in the catalogue.
| API-based MCPs (Composio, Zapier, content tools) | Local open-source servers | Hosted session services | Reduck | |
|---|---|---|---|---|
| People search and other profiles | No | Yes | Yes | Yes |
| Your inbox | No | Yes | Yes | Yes |
| Setup | OAuth to a LinkedIn app | Python, uvx or Docker on your machine | Connect your LinkedIn on their site | Stay signed in to LinkedIn in Chrome |
| Works in ChatGPT | Yes | Only if you host it | Yes | Yes |
| Where your LinkedIn session runs | Not used | Your machine, in its own browser | Their servers | Your own Chrome |
Google the hottest LinkedIn posts on "AI agents that browse the web", pull profiles of people who reacted/commented, and flag those matching "CTO building AI agents at a US startup".
Growth hackingLet Claude, ChatGPT or any AI agent export a LinkedIn Sales Navigator lead list or search to CSV or Google Sheets, every page, with each lead's name, title, company, location and Sales Navigator link, from your own Sales Navigator seat.
Growth hackingGive Claude, ChatGPT or any MCP client your X account, as you, with no developer app, no API key and no pay-per-use credits.
Social media