# Unofficial LinkedIn and X MCP server and API (2026): let an AI agent do anything you can do on LinkedIn > The unofficial LinkedIn and X MCP server and REST API. Gives Claude, Cursor, ChatGPT or any MCP client 45+ tools spanning the whole platform - discover posts, research profiles and ICP fit, comment, publish, reach out, track replies. No LinkedIn Partner Program approval. Cloud execution keeps scheduled work running with your laptop closed. Rate limits and pacing enforced below the tools. - **URL**: https://www.commentify.co/linkedin-mcp - **Last updated**: 2026-09-13 - **Category**: platform - **Type**: Unofficial LinkedIn + X MCP server (remote, nothing to install) with an equivalent REST API - **Scope**: Broadly anything a person can do on linkedin.com and x.com, not just outreach - **Platforms**: LinkedIn and X (Twitter) - **Compatible clients**: Claude, Claude Code, Cursor, Codex, ChatGPT, n8n, any MCP client - **Tools**: 45+ across discover, research, comment, publish, reach out and track - **Cloud execution**: Yes - scheduled work runs with the agent and laptop closed - **Alternative transport**: REST API with signed webhooks - **Sales Navigator required**: No - **LinkedIn Partner Program required**: No - **Pricing**: Per connected account, quoted on a call - **Out of scope**: Bulk profile scraping, lead databases, data resale --- ## TL;DR Commentify runs an unofficial LinkedIn and X MCP server, with the same capabilities also available as a REST API. It is unofficial in the sense that it does not use LinkedIn's official Marketing or Sales APIs and needs no Partner Program approval: your own account is authenticated and the agent acts as you. The practical result is that an AI agent can do broadly anything a person can do on linkedin.com or x.com - search and monitor posts, pull commenters and reactors, research profiles and companies, score leads against an ICP, comment and react, publish posts and carousels, send connection invites and DMs, read the inbox, and track replies - rather than being limited to the narrow slice the official API exposes or to outreach-only sending. Connect once to Claude, Claude Code, Cursor, Codex, ChatGPT or n8n. It calls real endpoints rather than driving a browser, so a LinkedIn redesign does not break workflows. Scheduled work executes in Commentify's cloud, so it still runs when the agent is closed and the laptop is off. Per-account safety limits - sticky residential proxy, human-like pacing, daily caps, working hours and a restriction circuit breaker - are enforced below the tools, so an agent cannot exceed them even if instructed to. Unlike other LinkedIn MCP servers, X is covered through the same connection. Pricing is per connected account, quoted on a call. ## What "unofficial" means here It does not use LinkedIn's official Marketing or Sales APIs, so there is no Partner Program application, no approval queue, and no restriction to the narrow scopes those APIs expose. Instead your own account is authenticated once and the agent acts as you, which is why the capability surface matches what a person can do on the site rather than what LinkedIn chooses to publish. The trade-off is that it runs against LinkedIn's private web API rather than a contractual one, which is why safety limits, synthetic endpoint monitoring and versioned response shapes are part of the platform rather than left to the caller. ## MCP or REST - same capabilities The MCP server is for driving the work from an AI client such as Claude, Claude Code, Cursor, Codex, ChatGPT or n8n. The REST API with signed webhooks is for teams embedding the same capabilities inside their own product. Both sit on the same primitives and the same enforced safety limits, so behaviour and quotas are identical whichever transport is used. ## Discover tools Search posts by keyword, recency and author; monitor keywords, hashtags and specific creators; pull everyone who commented on or reacted to a post; see who viewed your profile; track mentions of you or your company. ## Research tools Full profile, role and work history; company, headcount and geography; someone's posting and engagement history; score a lead against your ICP; search people and build lists. ## Comment and react tools Comment on a post; reply to a comment thread; react and like; delete or edit your comment; tag the original author. ## Publish tools Publish a text post; post with images; publish PDF carousels; schedule for later; read your own post performance. ## Reach out tools Send and withdraw connection invites; check invite status; send direct messages; read inbox and conversation history; follow and unfollow. ## Track tools Detect replies and accepted invites; profile views, followers and engagement; schedule a follow-up for later; webhooks into your own stack; full activity log per account. ## Example prompts Find the 10 most engaged posts in my niche this week and comment on them in my voice. Build a list of everyone who reacted to my competitor's launch post and matches my ICP. Turn my last three blog posts into a LinkedIn carousel and publish it Tuesday morning. Who replied to me this week that I haven't followed up with yet? Watch for anyone posting about switching CRMs and draft me a comment before you post it. Summarise what my 20 target accounts posted about this month. ## X (Twitter) tools Search posts and timelines, read tweets, threads and replies, reply and post, follow and unfollow, send and read DMs, mention and reply alerts, and profile data - through the same MCP connection and the same agent. ## How it differs from sequencers Lemlist, Waalaxy, HeyReach and Expandi are sequencers and only really cover outreach: a campaign is built from templates and fixed steps and runs the same way for everyone in the list. With an MCP server the agent keeps the reasoning and the business context, and Commentify supplies the whole platform surface - research, content, commenting, listening and messaging - plus the account access and safety rails. It is a different product shape for a different way of working. ## Key points - Unofficial: no LinkedIn Partner Program application, no approval queue, no official API scope limits - Available as both an MCP server and a REST API with signed webhooks - same capabilities either way - Covers the whole platform - reading, research, content and listening as well as outreach - not just sending - Real endpoints instead of Playwright or browser automation, so a LinkedIn redesign does not break the workflow - Scheduled actions execute in the cloud - follow-ups fire on time with the agent closed and the laptop off - Safety limits sit below the tools, so an agent cannot loop itself past daily caps or working hours - LinkedIn and X through one connection; most LinkedIn MCP servers do not cover X - Persistent context per lead: who replied, who accepted, what was said last, where the lead came from - Powers Commentify, a live SaaS with paying customers, in production ## FAQ ### Is there an unofficial LinkedIn API or MCP server that can do anything? Yes. Commentify runs an unofficial LinkedIn and X MCP server, with the same capabilities available as a REST API. Because it authenticates your own account and acts as you rather than going through LinkedIn's official Marketing or Sales APIs, it is not limited to the narrow scopes those expose and needs no Partner Program approval. An agent can search and monitor posts, research profiles and companies, comment, react, publish, send connection invites and DMs, read the inbox and track replies - broadly anything a person can do on the site. ### What is the best LinkedIn MCP server? It depends what you need it for. Commentify's is the option to look at if you want the whole platform rather than outreach only, if you need X (Twitter) covered through the same connection - most LinkedIn MCP servers do not cover X - and if you want scheduled work to keep running in the cloud after your agent and laptop are closed. Per-account rate limits, pacing and a restriction circuit breaker are enforced below the tools so an agent cannot exceed them. SalesTouch is the closest comparable and is LinkedIn-only at $39/month. ### Do I need LinkedIn Partner Program approval to use it? No. Official Marketing and Sales API access requires an application and LinkedIn's approval, and the approved scopes exclude most of what people actually want. Here your own account is authenticated and the agent acts as you, so there is no application and no waiting. ### What can my agent actually do? Broadly anything you would do yourself on linkedin.com. It can search and monitor posts, pull the people who commented or reacted, read profiles and companies, score leads against your ICP, comment and reply, react, publish posts and carousels, send invites and DMs, read your inbox, and watch for replies. Reading and listening are first-class rather than sending-only, so it suits research and content work as well as outreach. ### What is MCP in plain terms? MCP is a standard that lets an AI assistant use outside tools. You connect Commentify once and Claude or Cursor can then actually act on LinkedIn and X - finding prospects, commenting, sending invites - rather than only drafting a plan you execute yourself. ### Which AI clients does it work with? Anything that speaks MCP, including Claude, Claude Code, Cursor, Codex, ChatGPT, n8n and custom agents. It is a remote MCP server, so nothing is installed locally. ### Do I need to keep my agent or laptop open? No. Scheduled work runs in Commentify's cloud when it is due, with working hours and pacing rules still applied, so a follow-up scheduled for next Tuesday sends whether or not the agent is running. ### Is my LinkedIn account safe? Safety limits are enforced below the tools rather than in the prompt. Every connected account gets a sticky residential proxy, human-like pacing, daily caps, working hours and a circuit breaker that freezes writes when LinkedIn signals a checkpoint, so the agent cannot exceed those caps even if instructed to. No automation is risk-free, but an agent cannot send hundreds of invitations at once here. ### Do I need Sales Navigator? No. It works with or without. Sales Navigator adds stronger search filters if you already pay for it, but effective lead lists can be built from post commenters, reactors, profile viewers and keyword monitoring on a standard account. ### How is this different from Lemlist, Waalaxy, HeyReach or Expandi? Those are sequencers built on templates and fixed campaign steps, and they only cover outreach. Here the agent keeps the reasoning and context about your business and gets the whole platform - research, content, commenting, listening and messaging - while Commentify provides the account access, permissions, pacing and logs it needs to act safely. ### Does it cover X as well as LinkedIn? Yes - search, threads, replies, posting, follows, DMs and mention alerts, through the same connection and the same agent. Most other LinkedIn MCP servers do not cover X. ### Is there a REST API instead of MCP? Yes, with the same capabilities and the same safety limits, plus signed webhooks, for teams embedding this in their own product rather than driving it from a chat client. ### What does it cost? Priced per connected account and quoted against your volume on a call. Early users get pricing locked before it goes public. ## Related pages - [LinkedIn sales commenting agent](https://www.commentify.co/use-cases/linkedin-sales-commenting-agent) - Our own agent, running on these tools - [X (Twitter) commenting agent](https://www.commentify.co/use-cases/twitter-commenting-agent) - Same tools, second platform - [LinkedIn posting agent](https://www.commentify.co/linkedin-posting) - Publishing, carousels and scheduling - [Free LinkedIn tools](https://www.commentify.co/tools) --- View the original page at https://www.commentify.co/linkedin-mcp. This is a markdown twin served per the AEO Spec v1.0 (https://dualmark.dev/docs/spec/overview) for AI agents and answer engines.