The best LinkedIn MCP servers (2026): which ones find leads, which ones send, and what is official

TL;DR: The best LinkedIn MCP server for Claude, ChatGPT and other AI chats is traxy if you want to run LinkedIn GTM end to end from chat: it finds high intent leads from social intent, researches each one, and runs LinkedIn and email campaigns and inbox replies from your own accounts, with your yes in the chat before anything sends. Every traxy customer can run the whole campaign from Claude, ChatGPT, Codex, Cursor or Grok: once you approve it, the campaign runs on autopilot from your own LinkedIn and email accounts, or you approve every message. Most other LinkedIn MCP servers send messages or pull profiles: proReach, ReachHunt and HeyReach run LinkedIn campaigns, LinkedNav and BeReach add signal-based lead finding to LinkedIn sending, Periodix and Linked API give an agent single LinkedIn actions, and open-source servers drive your own browser session. LinkedIn has no official MCP server for outreach, so every option here is third-party.

Checked November 10, 2026. traxy wrote this comparison and is one of the vendors in it. Every other server is described only from its own website, docs or GitHub page, linked where it appears and checked November 2026. MCP servers change often, so confirm details with each vendor before you connect one.

Which LinkedIn MCP server is best for research, posting, data extraction or outreach?

It depends on the job: stickerdaniel/linkedin-mcp-server for research on your own LinkedIn, Composio for posting, Apify or Bright Data for data extraction, Unipile for building LinkedIn and email features into your own product, and traxy to find high intent leads and run the outreach to them from one connection. traxy runs your entire LinkedIn GTM end to end, from finding high intent leads from social intent to enriching and researching each one, to running relevant, personalized LinkedIn and email outreach to them on autopilot. The campaign runs on its own only after you approve it, or you approve every message.

Last checked November 11, 2026. traxy wrote this table and is one of the vendors in it. Every other line comes from that vendor's own page or GitHub repo, linked in its row.

If you only need research, posting or bulk data extraction, the server built for that job fits better than traxy. LinkedNav and BeReach also pair lead finding with LinkedIn sending, and the comparison below shows how they differ on research, email and approval. traxy is the pick when you want the people to contact and the LinkedIn and email outreach to them in one connection. See LinkedIn outbound with traxy for how the campaigns run.

Is there a LinkedIn MCP server for outreach?

Yes, several, and all of them come from third parties. LinkedIn doesn't publish an MCP server for prospecting or outreach. Its Hiring Pro connector in Claude, announced October 21, 2026, covers Hiring Pro job posts only, and its open API permissions let any developer sign members in and post, comment and like as them, but not send messages or connection requests. Our guide Is there a LinkedIn API for outreach? covers what LinkedIn's own APIs allow and what LinkedIn says about automation tools.

The LinkedIn MCP servers for outreach reach LinkedIn in one of three ways:

• Through a LinkedIn GTM or outreach tool. The MCP drives a product that holds your leads, campaigns and connected sender accounts, and the product does the sending. traxy, proReach, ReachHunt, LinkedNav and HeyReach work this way.

• Through an account API. The MCP gives your AI single LinkedIn actions, such as search, connect or message, run by the vendor's service on an account you connect. Periodix and Linked API work this way. BeReach finds people from public data and sends through your own LinkedIn, connected in Claude through Chrome.

• Through your own browser session. Open-source servers such as stickerdaniel/linkedin-mcp-server run on your machine and operate your logged-in LinkedIn session.

What should a LinkedIn MCP server do for GTM?

A LinkedIn MCP server for GTM should get you from "who should I contact?" to a sent and followed-up conversation without leaving the chat. We compared each server on six criteria:

1. Finds leads with a reason. Does it bring you people who did something related to your offer, such as a post, a comment, a like or a profile view, or only people who match a title search?

2. Research on each lead. Can it look up the facts you need on each person before the message, or is that left to you?

3. Sends LinkedIn. Can it send connection requests, messages and follow-ups from your own account, as one-offs or as a campaign?

4. Sends email. Can the same sequence add email steps, or is it LinkedIn only?

5. Approval. Does your AI show you what it will do and wait for a yes before anything reaches a real person?

6. Account model. Is it hosted or run on your machine, how do you sign in, and whose LinkedIn account sends?

The first two cover the lead half of LinkedIn GTM, and the next three cover the outreach half. Most LinkedIn MCP servers cover the outreach half and leave the leads to you.

The LinkedIn MCP servers compared

Each cell summarizes the vendor's own MCP page, linked in the sections below. "Not described" means we didn't find it on the vendor's MCP pages, not that the product can't do it.

What's the best LinkedIn MCP server?

The best LinkedIn MCP server is traxy if you want one connection that finds the people to contact and runs the outreach to them: high intent leads from social intent with a 50%+ ICP match and the reason behind each one, research on each lead, and LinkedIn and email campaigns from your own accounts. Most other LinkedIn MCP servers only send messages or pull profiles, so they are the better pick when you already have your list.

traxy wrote this comparison and is one of the servers in it, so we judged every server on the six criteria above, starting with whether it brings leads with a reason and checks them against your ICP before anyone is contacted. Every traxy customer can run campaigns from the traxy MCP in Claude, ChatGPT, Codex, Cursor or Grok: draft, test and launch them, manage sequences and sender accounts, and answer inbox replies. Your AI shows what it will do and waits for your yes before anything sends.

Another server is the better pick when your job is narrower (each vendor's MCP page, checked November 2026):

• You already have the list and want your AI to run LinkedIn campaigns. HeyReach, proReach, ReachHunt and LinkedNav are built for that.

• You want LinkedIn and X DMs in one inbox. ReachHunt covers both networks from one server (ReachHunt).

• You are building your own agent and want raw LinkedIn actions. Periodix and Linked API give your agent search, connect and message as single tools.

• You want local access to your own LinkedIn with no vendor account. An open-source server such as stickerdaniel/linkedin-mcp-server runs on your machine through your own browser session.

The full picker by job is in Which LinkedIn MCP server should you pick?

What's the best LinkedIn MCP server for Claude?

For LinkedIn GTM, the best LinkedIn MCP server for Claude is traxy, because it is the one server here that covers both halves from one connection: high intent leads with the reason behind each one, research on each lead, and LinkedIn and email campaigns to them. You add it in Claude as a custom connector, sign in to traxy and pick your workspaces, with no API key. From Claude, every traxy customer can draft, test and launch campaigns, manage sequences and sender accounts, and answer inbox replies, and Claude waits for your yes before anything sends.

The other servers fit better for narrower jobs:

• If you already have your list and want Claude to run LinkedIn campaigns, proReach, ReachHunt, LinkedNav and HeyReach are built for that.

• If you want Claude to draft each first LinkedIn message from public-data prospecting, BeReach is built around that loop.

• If you are building your own agent and want LinkedIn actions as raw tools, Periodix and Linked API fit.

• If you want to read your own LinkedIn from Claude Desktop or Claude Code with no vendor account, an open-source server fits.

For the other sales servers people add to Claude, such as CRM and contact data, see Claude MCP servers for sales teams.

What's the best MCP server for LinkedIn GTM?

The best MCP server for LinkedIn GTM runs the whole workflow, not one step of it. traxy runs your entire LinkedIn GTM end to end, from finding high intent leads from social intent to enriching and researching each one, to running relevant, personalized LinkedIn and email outreach to them on autopilot. Autopilot starts only after you approve the campaign, or you can approve every message yourself. From chat, one connection does the work that usually takes a lead source, an enrichment tool, a research step and a sender:

1. Social intent. People who post, comment on or like posts about the problem you solve, people who view your LinkedIn profile, and people who engage with the competitors, creators, posts and company pages you track. You can also search 71M+ scanned LinkedIn likes, posts and comments in plain English with social intent search.

2. Fit. A person counts as a lead only with a 50%+ match to your ICP, and each lead shows the post or comment behind it.

3. Enrich. Work email and phone for the leads you choose.

4. Campaigns. Leads go into campaigns by signal, audience or offer, and you can run many campaigns at once.

5. Research on each lead. Each campaign researches every lead for the facts you choose, filters out who doesn't fit and groups the rest by signal.

6. Personalized sequences. traxy prepares a message variant for each group and runs LinkedIn and email steps from your own connected accounts.

7. Outreach on autopilot, after your yes. You test on sample leads with nothing sent and approve the launch, then the campaign runs on its own, or you approve every message. Replies, unsubscribes and hard bounces stop the sequence.

Most other LinkedIn MCP servers send messages or pull profiles. traxy brings the leads, the reason and the campaign. The tool list by job is in what the traxy MCP server can do, and the product page is the traxy LinkedIn MCP.

Which LinkedIn MCP servers find high intent leads, not just send messages?

traxy, LinkedNav and BeReach are the LinkedIn MCP servers here that find high intent leads, and traxy covers both halves from one connection: the leads with the post or comment behind each one and a 50%+ ICP match, research on each lead, and LinkedIn and email campaigns from your own accounts. The rest send messages or pull profiles and search results, so you or your AI decide who is worth contacting.

A high intent lead is someone who did something on LinkedIn tied to what you sell, such as a post, a comment, a like, a profile view or engaging with a post you track, and who fits your ICP. A title search gives you a list, not a reason to reach out. Here is what each server does, from its own MCP page (checked November 2026):

• traxy. Signals and Tracking turn LinkedIn activity into leads: people who post, comment on or like posts about the problem you solve, view your LinkedIn profile, or engage with the competitors, creators, posts and company pages you track. Only people with a 50%+ ICP match count as leads, each with the post or comment behind it, and you can search 71M+ scanned LinkedIn likes, posts and comments in plain English from chat. Then traxy runs the campaign to them, after your yes (traxy LinkedIn MCP).

• LinkedNav. Signal leads from its intent, competitor and influencer agents, filtered by score, and the engagers of monitored posts scored against your ICP (LinkedNav).

• BeReach. Each lead comes with a dated reason to write, such as a post, a new job or a raise, and its tools include a fit check (BeReach).

• Engagers or search, with the fit check left to you. ReachHunt searches LinkedIn by keyword, title, company, location and industry and imports post likers, commenters and group members. Periodix lists the people who reacted to or commented on a post. proReach turns a plain-language ICP description into a LinkedIn search. Linked API and open-source servers such as stickerdaniel/linkedin-mcp-server search people, companies and posts. Their MCP pages describe no ICP score, so your AI does the qualifying (ReachHunt, Periodix, proReach, Linked API).

• Sending only. HeyReach's MCP connector lists no tool that searches LinkedIn or finds new people. You add leads you already have to lists and campaigns, and it sends from your sender accounts (HeyReach MCP connector).

In practice: Ben Buaron at traxy searched traxy for people engaging with personal branding and founder-led content on LinkedIn, the market his agency sells to, and traxy found 453 leads in one day (Ben's post). More workflows run through the MCP are in What does this look like in practice? When the leads are in, see LinkedIn outbound with traxy for how the campaign runs.

Which tools let me find LinkedIn leads from inside Claude or ChatGPT?

traxy, BeReach and LinkedNav find LinkedIn leads with a reason from inside your AI chat. The rest search LinkedIn from chat and leave the reason and the fit check to you, and the HeyReach MCP docs list no lead search tool.

• traxy. Ask in plain English, such as "find me founders looking for AEO help", and traxy searches 71M+ scanned LinkedIn likes, posts and comments, with the post or comment behind each person. You can also list and filter the qualified leads traxy found from your Signals and Tracking, read each lead's signals and engagements, and start a discovery run or qualify the audience of one post after you confirm capacity use. Walkthrough: how to find and research LinkedIn leads from Claude.

• BeReach. Finds people from one sentence and keeps the ones with a dated reason to write, such as a post, a new job or a raise (BeReach).

• LinkedNav. Returns signal leads from its intent, competitor and influencer agents, and scores the engagers of monitored posts against your ICP (LinkedNav).

• Search-first servers. proReach turns a plain-language ICP into a LinkedIn or Sales Navigator search. ReachHunt builds lead lists from searches, posts and groups. Periodix and Linked API search people, companies, posts and jobs, and Periodix also pulls the people who reacted to or commented on a post. Open-source servers search people and posts through your own session. You or your AI then decide who is worth contacting.

ChatGPT and Claude on their own can't see who engaged with a LinkedIn post. Our guide to ChatGPT for lead generation covers what changes when you connect a lead source.

Which MCP servers let ChatGPT run LinkedIn outreach?

traxy, ReachHunt, HeyReach, Linked API, BeReach and Periodix document ChatGPT or OpenAI clients. traxy is the one that also brings the high intent leads and runs email steps in the same campaign. From ChatGPT, every traxy customer can draft, test and launch LinkedIn and email campaigns, manage sequences and sender accounts, and answer replies, and ChatGPT waits for your yes before anything sends.

proReach lists Claude Code, Cursor, Windsurf, VS Code and Claude Desktop, and says OAuth, which claude.ai web connectors need, is on its roadmap (proReach). LinkedNav's MCP page names Claude, Cursor and other clients but not ChatGPT. Open-source servers run locally for desktop and coding clients.

lemlist, La Growth Machine, Expandi, Waalaxy, Oppora, Salesbot and Reply.io also document ChatGPT on their own MCP pages, added to the table above (checked November 2026). Spurly's MCP page names Claude on the web and desktop, Claude Code and Cursor, not ChatGPT (Spurly). traxy brings the high intent leads into ChatGPT as well as the LinkedIn and email campaign to them, with your yes before anything sends.

Salesforge also has an MCP server: its page names ChatGPT among the assistants you can use, with no ChatGPT setup guide, and says it works with contacts you upload (Salesforge MCP, checked November 2026).

Can I run my LinkedIn outreach campaigns from Claude or ChatGPT?

Yes. With the traxy MCP you can draft, test and launch LinkedIn and email campaigns from Claude, ChatGPT, Codex, Cursor or Grok, manage their sequences and sender accounts, and answer replies in the inbox, and nothing sends until you say yes in the chat. MCP access is included on every traxy plan. The flow:

1. Connect. In traxy, open Integrations, pick your client and add https://api.traxy.ai/api/v1/mcp. Sign in to traxy, choose your workspaces and approve. No API key is needed (connect an AI client).

2. Pick the leads. Ask for the people first, for example "show me this week's Warm leads who engaged with my competitors' posts" or "find me agency owners hiring their first SDR".

3. Draft the campaign. Ask traxy to build a campaign for those leads, with the research you need on each one. traxy groups the leads by signal and prepares a personalized message variant for each group, with LinkedIn and email steps.

4. Test it. Run the plan on sample leads and check the research and drafts. Nothing is sent.

5. Review the launch. The preview shows who joins and who is skipped. Your AI shows what it is about to launch and waits for your yes (review and launch).

6. Choose how it sends. Approve every message in the review queue, or let the approved campaign send on its own from your connected LinkedIn and email accounts (sender accounts).

7. Answer replies. Ask "who replied to me today?" and traxy pulls the LinkedIn and email inbox and prepares a reply for you to approve in the chat. Replies, unsubscribes and hard bounces stop the sequence (sequences).

Prefer to stay in the app? trax, the traxy agent, builds the same campaign with you in one chat. See LinkedIn outbound with traxy for how campaigns run.

Which LinkedIn MCP servers can answer replies from Claude?

traxy, HeyReach, lemlist, La Growth Machine, Salesbot, Reply.io and the open-source stickerdaniel/linkedin-mcp-server can send a reply from Claude, while Expandi drafts the reply for you to approve, Spurly does the same unless you turn on two switches, and Oppora's MCP can't read reply text at all. With traxy, the same connection also finds the high intent leads you are talking to, with a 50%+ ICP match and the post or comment behind each one, and runs the LinkedIn and email campaign that started the conversation; Claude shows each reply and waits for your yes before it sends.

Checked November 11, 2026. traxy wrote this table and is one of the vendors in it. Every other row comes from that vendor's own MCP page, docs or GitHub repo, linked in the row.

Of the servers compared above, ReachHunt's MCP answers the inbox, LinkedNav's AI reply drafts wait for your approval unless you allow auto-send on routine threads, and proReach's inbox tool is read-only, so a person sends every reply. LinkedIn's own Messages API, which lets members message first-degree connections or reply to existing conversations, is restricted to approved partners (LinkedIn Messages API docs, checked November 2026).

Salesforge's MCP also reads new replies and answers them on email and LinkedIn from Claude, Cursor or any MCP client; whether the assistant asks before a reply goes out is not stated on its page (Salesforge MCP, checked November 2026).

In practice: on November 9, Ben Buaron at traxy showed the flow from Claude, Codex or Grok. He asks "who replied to me today?", and the traxy MCP pulls the LinkedIn and email inbox, prepares each reply, and sends it once he approves it in the chat (Ben's post).

Where another server fits better: if your campaigns already run in HeyReach, lemlist or La Growth Machine and you only want to clear that inbox from chat, their own MCP does it without moving tools. If you want every reply to wait in the vendor's app for a click, Expandi and Salesbot work that way by default. traxy fits when you want the leads, the campaign and the replies in one connection. See LinkedIn outbound with traxy for how the campaigns and inbox run.

Can Claude launch a LinkedIn and email campaign, or only add leads to one?

Claude can launch a campaign with LinkedIn and email steps through traxy, lemlist, La Growth Machine, Oppora and Reply.io, and a LinkedIn campaign through HeyReach and Salesbot; Waalaxy's MCP only adds prospects to a campaign you already set up, and Expandi builds the draft but you launch it yourself. With traxy you test the campaign on sample leads with nothing sent, review who joins and who is skipped, and say yes in the chat; then it runs on autopilot from your own LinkedIn and email accounts, or you approve every message.

Checked November 11, 2026. traxy wrote this table and is one of the vendors in it. Every other row comes from that vendor's own MCP page or docs, linked in the row.

Of the servers compared above, proReach launches behind a confirmation and its own worker performs the LinkedIn actions later inside daily caps, working hours and cooldowns, and Periodix builds multi-step sequences that enroll up to 500 recipients across your connected profiles.

Salesforge's MCP can also launch a campaign with email and LinkedIn steps from Claude, Cursor or any MCP client, for contacts you upload; its FAQ says it doesn't find new leads (Salesforge MCP, checked November 2026).

Where another server fits better: if your campaigns already live in lemlist, La Growth Machine or HeyReach and you only want to start and pause them from chat, use that tool's own MCP. If you want an AI that can't start a campaign at all, Expandi drafts it for you to launch and Waalaxy only adds prospects to campaigns you set up. traxy fits when the same chat should find the high intent leads, test the campaign and launch it. The full flow is in Can I run my LinkedIn outreach campaigns from Claude or ChatGPT?

Where Salesforge fits better: if your team already runs its outbound on Salesforge and uploads its own lists, its MCP launches those campaigns and answers the replies without moving tools. traxy fits when you also want the high intent leads, with the LinkedIn and email campaign to them and a yes in chat before anything sends (LinkedIn outbound with traxy).

Which LinkedIn MCP servers can read who engaged with a post?

traxy, La Growth Machine, Oppora and Expandi can pull the people who engaged with LinkedIn posts beyond your own, and traxy keeps only the engagers with a 50%+ ICP match, each with the post or comment behind them, then runs the LinkedIn and email outreach to them. Postiv reads comment threads only on posts in your own Postiv workspace, and vlpmedialtd/linkedin-mcp, stickerdaniel/linkedin-mcp-server and the HeyReach and Waalaxy MCPs list no tool that returns a post's engagers.

This is the deciding question for lead finding: the people who like, comment on or react to a post about your problem are the high intent leads, and a server that can't read them leaves you with a title search. With traxy you paste any LinkedIn post into your AI chat, yours, a competitor's or an influencer's, ask for its audience and confirm the capacity use, and you get back the engagers who fit your ICP (how to find and research LinkedIn leads from Claude).

Checked November 11, 2026. traxy wrote this table and is one of the vendors in it. Every other row comes from that vendor's own MCP page, docs or GitHub repo, linked in the row.

Of the servers compared above, Periodix pulls the people who reacted to or commented on a post, BeReach reads engagers and finds work emails, ReachHunt imports post likers and commenters, and LinkedNav scores the engagers of monitored posts against your ICP.

Where another server fits better: if you mainly want to read and answer comments on your own company's posts as a content team, Postiv is built for that workflow. If your lists already come from La Growth Machine or Oppora searches, their engager tools sit next to the rest of your list building. traxy fits when you want the engagers checked against your ICP and the outreach to them run from the same chat.

MCP servers that find the people engaging with a LinkedIn post

SyncGTM, Unipile, Crispy and Apify actors can return the people who reacted to or commented on a LinkedIn post, but none of their pages describes checking those people against your ICP, and most leave the outreach to another tool. traxy qualifies a post's audience against your ICP, shows the post or comment behind each lead, and then runs LinkedIn and email outreach to them in the app, through trax or from the MCP after your yes in chat.

A list of everyone who liked a post is mostly people who will never buy from you. The fit check and the reason to write are what turn it into high intent leads, and that is the step these tools hand back to you or your AI.

Checked November 11, 2026. traxy wrote this section and is one of the vendors in it. Every other line comes from that vendor's own website, docs or tool page, linked where it appears.

In practice: Vamo turned engagement on 64 posts it had already published into 14 replies and 11 meetings, with traxy flagging the engagers who fit its ICP (Vamo case study).

Where another server fits better: SyncGTM if you want raw reactor and commenter rows plus email lookups to feed a sequencer you already run. Unipile if you are building engager features into your own product. Crispy if you want a LinkedIn-only tool on your own account with post, comment and inbox tools in one MCP. An Apify actor if you want a bulk export of a post's engagers. traxy fits when you want only the engagers who fit, with the post or comment behind each, and the LinkedIn and email outreach to them in the same place.

Which LinkedIn MCP servers find leads, which send, and which do both?

Of the other servers buyers ask about, SyncGTM, Datamagnet, Manifold and Apify actors find people from LinkedIn activity but don't send; HeyReach, lemlist, Sliq, LinkMCP and Unipile send, or let you build sending, to people you bring or search for; and Valley, FirstTouch and PumpGTM do both. traxy does both from one connection: high intent leads from social intent with a 50%+ ICP match and the post or comment behind each one, then LinkedIn and email outreach to them from your own connected accounts.

Salesforge belongs with the senders: its MCP works with contacts you upload, and lead search is a separate product (Salesforge MCP, checked November 2026).

Find leads from LinkedIn activity, no sending

• SyncGTM. Returns a post's reactors and commenters over MCP. Emails come from a separate find_work_email step, and its pages don't describe sending (SyncGTM tool docs). Fits better if you want engager rows and email lookups for a sequencer you already use.

• Datamagnet. Its MCP server reads LinkedIn people, companies, posts and jobs, lists who engaged with a post through get_post_engagement, and has no messaging tool. You sign in to Datamagnet, not to LinkedIn (Datamagnet MCP). Fits better if you want LinkedIn data in your AI with no LinkedIn account connected.

• Manifold. Reads public LinkedIn posts, profiles, company pages and a post's commenters through a data provider, without signing in to LinkedIn, and its people tools find and verify work emails. Its LinkedIn page describes no tool that messages anyone (Manifold LinkedIn MCP). Fits better if you want public LinkedIn reads and email finding next to other marketing data.

• Apify actors. Store Actors export a post's likers, commenters and resharers, run from your AI through the Apify MCP server (LinkedIn Posts Engagers). Fits better for bulk exports you process yourself.

Send to leads you bring or search for

• HeyReach. Its MCP connector lists no lead search tool. send_message and start_campaign are marked as writes that reach LinkedIn, and every tool says whether it reads or changes data, so your assistant can ask you first. Claude Code connects with sign-in; Cursor connects with an MCP API key (HeyReach MCP connector). Fits better if you run many LinkedIn sender accounts, for example as an agency, and already have your lists.

• lemlist. Searches its 450M+ B2B database, enriches leads, and creates and runs campaigns. Tools that send, delete, spend credits or change campaigns require your confirmation. Its MCP guide doesn't cover finding people from LinkedIn post activity (lemlist MCP guide). Fits better if you want campaigns built on a large B2B contact database.

• Sliq. Searches people and companies across LinkedIn, Apollo and other sources, watches prospects' posts, and runs LinkedIn and email outreach. By default every outreach message waits in its approval queue, but connection requests without a note send automatically unless you turn on approval for them. Claude Code connects with a Sliq API key in a header (Sliq MCP). Fits better if you want to choose, per channel and step, what sends on its own and what asks first.

• LinkMCP. You connect your own LinkedIn once in its web app. It reads profiles, messages and recent posts, sends connection requests and messages within daily limits (20 connection requests and 25 messages a day on its default Cautious tier), and finds work emails. LinkMCP says it isn't affiliated with or endorsed by LinkedIn and doesn't use an official LinkedIn partner API, and it suggests asking for drafts and telling your assistant not to send anything until you have reviewed them (LinkMCP). Fits better if you want your own LinkedIn in Claude, ChatGPT or Cursor one action at a time.

• Unipile. A building block, not a finished outreach tool: its MCP helps a coding agent build LinkedIn and email features into your product, which then calls Unipile's API (Unipile MCP). Fits better if you are building your own product.

• Salesforge. Its MCP FAQ answers whether it can find new leads with "No. It works with contacts you upload." To search for new leads, you connect Leadsforge with its own API key. From Claude, Cursor or any MCP client you upload contacts one by one or in bulk, launch campaigns with email and LinkedIn steps, and answer email and LinkedIn replies. Whether the assistant asks before sending is not stated on its page (Salesforge MCP). Fits better if your team already runs its outbound on Salesforge and uploads its own lists.

Find leads and send

• Valley. Its post-engager page says it captures everyone who engaged with your posts, scores each one against your ICP, researches the ones that fit and sends a personal message across LinkedIn and email, and that you approve every send. Its MCP gives Claude, ChatGPT and Cursor your campaigns, analytics, inbox, prospects and lists; every action that changes data is marked for approval, and sends are confirmed twice (Valley post engagers, Valley MCP). Fits better if the engagers of your own posts are your main lead source and you want two confirmations on every send.

• FirstTouch. Its MCP finds and enriches prospects, watches who likes and comments on your team's posts, qualifies them against your ICP, and sends connection requests, messages and InMails, with email steps configured inside its flows. Approvals are on by default and configurable by action type, and every social touch logs to HubSpot (FirstTouch, FirstTouch MCP). Fits better if your team works in HubSpot and wants the engagers of its own posts routed to the right teammate.

• PumpGTM. Its Claude Code page shows finding buyers from hiring signals, competitor case studies and LinkedIn activity, such as watching a founder's posts and messaging the people who comment, then sending LinkedIn and email from your own accounts in the same sequence. It says nothing is sent until you approve, and follow-ups stop when someone replies. Its docs add X outreach (PumpGTM for Claude Code, PumpGTM docs). Fits better if you want LinkedIn, email and X from one server, or hiring signals as a main lead source.

Where traxy fits

traxy runs your entire LinkedIn GTM end to end, from finding high intent leads from social intent to enriching and researching each one, to running relevant, personalized LinkedIn and email outreach to them on autopilot. Autopilot starts only after you approve the campaign, or you approve every message, and from your AI chat nothing sends until you say yes.

It sits in the "find leads and send" group, and the difference is the lead half. traxy finds people from what they post, comment and like on LinkedIn, from your profile views, and from the engagers of the competitors, creators, posts and company pages you track, not only your own posts. You can also search 71M+ scanned LinkedIn likes, posts and comments in plain English. Only people with a 50%+ ICP match become leads, each campaign researches every lead for the facts you choose, and you run it by hand in the app, through trax, or from the MCP in Claude, ChatGPT, Codex, Cursor or Grok. See LinkedIn outbound with traxy for how the campaigns run.

traxy is not the pick if you need X outreach next to LinkedIn, where PumpGTM fits, or a raw LinkedIn data feed with no outreach, where SyncGTM, Datamagnet or Manifold fit.

Checked November 11, 2026. traxy wrote this section and is one of the vendors in it. Every other line comes from that vendor's own website, docs or tool page, linked where it appears.

How do I run LinkedIn outbound from Claude Code or Cursor?

Add a LinkedIn MCP server's remote URL to Claude Code or Cursor, sign in, and ask for the leads and the campaign in the chat. With traxy the URL is https://api.traxy.ai/api/v1/mcp, sign-in is OAuth, and every traxy customer can find high intent leads, build, test and launch LinkedIn and email campaigns, manage sequences and senders, and answer replies from there, with a yes in the chat before anything sends.

Setting up traxy, from the traxy docs:

1. Open the guide in traxy. Open Integrations and pick your client. For Claude Code, use the terminal setup shown on traxy's Claude integration page; traxy also has a guide for Cursor, alongside Codex, Grok, OpenCode, Pi and Ollama (connect Claude, connect an AI client).

2. Add the endpoint. The remote MCP endpoint is https://api.traxy.ai/api/v1/mcp.

3. Sign in with OAuth. A traxy REST API key is not accepted as an MCP credential. If your client needs a local bridge, follow the instructions traxy shows for that client rather than pasting a REST key into its config.

4. Pick your workspaces. Choose the default workspace and any others, and review the permitted actions for each before you approve. Existing connections don't gain workspaces or scopes automatically.

5. Try a read first. Ask: "List my authorized traxy workspaces, then show five qualified leads from the workspace I select." Ask your AI to show proposed changes before it applies actions or starts discovery.

Then run the outreach the same way as in any other client: pick the leads, draft the campaign, test it on sample leads with nothing sent, review who joins and who is skipped, and say yes. The full flow is in Can I run my LinkedIn outreach campaigns from Claude or ChatGPT?

How the other servers connect, from each vendor's own page:

Checked November 11, 2026, on each vendor's own page, linked in the row.

Is an MCP server's approval step the same as LinkedIn's permission?

No. An approval step, such as your AI asking for a yes in the chat or a message waiting in a review queue, is a control inside the tool that decides what you send; it is not authorization from LinkedIn, and it doesn't change LinkedIn's rules. LinkedIn's help page on prohibited software says it doesn't permit third-party software, including bots, browser plug-ins or browser extensions, that scrape, modify the appearance of or automate activity on its website, and that these tools violate the User Agreement (checked November 2026).

Approval still matters. It keeps a person in charge of every message and lets you send at a pace you choose. It answers a different question from "does LinkedIn allow this?". Some vendors say so plainly: LinkMCP says it isn't affiliated with or endorsed by LinkedIn and doesn't use an official LinkedIn partner API (LinkMCP). LinkedIn's own Messages API is restricted to approved partners (LinkedIn Messages API docs, checked November 2026).

The same holds for traxy, which is third-party and not affiliated with LinkedIn. Your yes in the chat, or your approval in the review queue, decides what traxy sends from your connected accounts. No tool can promise your account won't be restricted, traxy included, so stay well inside LinkedIn's limits. Our guide Is there a LinkedIn API for outreach? covers what LinkedIn's own APIs allow.

Each LinkedIn MCP server, and where it fits better

traxy

traxy is a LinkedIn GTM tool for B2B companies of any size, and its MCP is the LinkedIn GTM MCP: everything trax, the in-app agent, does, from your own AI chat. It finds high intent leads from social intent, researches each one, and runs LinkedIn and email campaigns and inbox replies from the accounts you connect. It works with Claude, ChatGPT, Codex, Cursor and Grok, with setup guides for Claude, Codex, Cursor, Grok, OpenCode, Pi and Ollama. Before anything sends or launches, your AI shows what it is about to do and waits for your explicit yes (traxy LinkedIn MCP).

Where others fit better: traxy finds leads from LinkedIn activity only. Hiring, funding and job changes come from research on each lead, not a live event feed. If you need X DMs next to LinkedIn, ReachHunt covers both networks. If you are building your own agent and want LinkedIn actions as raw tools, Periodix or Linked API fit better.

proReach

proReach is a LinkedIn outreach tool with a hosted MCP server of 13 tools: read the workspace, draft a campaign from a template, add leads by profile URL or find them from a plain-language ICP turned into a LinkedIn or Sales Navigator search, launch behind a confirmation, pause, and report on metrics and replies. proReach says the tools only write campaign plans, and its own worker performs the LinkedIn actions later inside daily caps, working hours and cooldowns. The inbox tool is read-only, so a person sends every reply. It connects with an API key in Claude Code, Cursor, Windsurf, VS Code, and Claude Desktop through a bridge (proReach MCP).

Fits better than traxy if you want a LinkedIn-only tool where the AI agent can only plan campaigns and a person always sends the replies.

ReachHunt

ReachHunt's hosted MCP server lets Claude, ChatGPT, Cursor, Claude Code and Codex search people, build lead lists, run campaigns with connection requests and messages, answer the inbox and manage ReachHunt's AI agents that answer replies, on LinkedIn and X (Twitter). Its campaigns also send InMails. You sign in with OAuth, and actions go through the LinkedIn accounts you connect to ReachHunt with its browser extension. Tools that reach real people are marked so your AI app can ask you first, and the AI can't delete anything (ReachHunt MCP, MCP docs, API page).

Fits better than traxy if you want LinkedIn and X DMs in one MCP, one inbox and one set of campaigns.

LinkedNav

LinkedNav's MCP server has 145 tools across campaigns, contacts and lists, people and company search, signal leads, social listening, email and phone enrichment, a unified inbox with approvals, sender management and account audits. Its signal leads come from intent, competitor and influencer agents, and it scores the engagers of monitored posts against your ICP. You connect by pasting an API key on LinkedNav's connect page, and senders are connected with its Chrome extension and a dedicated proxy. LinkedNav says your client asks you to confirm before anything sends, and AI reply drafts wait for your approval unless you allow auto-send on routine threads (LinkedNav MCP, LinkedNav agent guide).

Fits better than traxy if you want one MCP with a very large LinkedIn outreach tool surface across many senders, including audits that diagnose why a campaign stopped sending.

BeReach

BeReach is a hosted LinkedIn MCP server with 33 tools. From one sentence it finds people and companies, keeps the ones with a dated reason to write, such as a post, a new job or a raise, reads profiles, posts and engagers, checks fit, finds work emails and phone numbers, and drafts a first message for each person. BeReach says finding, reading and drafting work on public data, and sending invitations and DMs uses your own LinkedIn, connected in Claude through Chrome. Drafts wait for your approval, and BeReach says you decide how much to hand over, from one message at a time to a morning routine (BeReach MCP, BeReach's guide).

Fits better than traxy if you want Claude to write each first LinkedIn message itself from public-data prospecting, and you don't need email sequences.

Periodix

Periodix Actions has a hosted MCP server with 16 tools for LinkedIn and Sales Navigator: search people, companies, posts and jobs, read full profiles, companies and recent posts, pull the people who reacted to or commented on a post, send connection requests, messages and InMail, and build multi-step sequences that enroll up to 500 recipients across your connected profiles. It connects with an API key, and you connect LinkedIn profiles in its dashboard. Its MCP page doesn't describe a confirmation step before sends (Periodix MCP docs).

Fits better than traxy if you are building your own agent and want LinkedIn and Sales Navigator actions, InMail included, as typed tools.

Linked API

Linked API's MCP server connects your LinkedIn account to Claude, Claude Code, ChatGPT, Codex, Cursor, VS Code and Windsurf. Your assistant can search people, companies, posts and jobs, read profiles, the feed and the inbox, message and connect, and post, comment and react, with Sales Navigator actions too. Linked API says every action runs in a dedicated cloud browser on your account and that it never stores your password. You authenticate with tokens in the connector URL, and each LinkedIn account gets its own MCP instance (Linked API MCP, installation).

Fits better than traxy if you want an assistant that acts as you across most of LinkedIn, posting and commenting included, one action at a time.

HeyReach

HeyReach's MCP connector lets Claude, ChatGPT and Codex run its LinkedIn outreach: create, update, start and pause campaigns, add leads to lists and campaigns, read conversations and send LinkedIn messages, manage tags, the blacklist, webhooks and voice-note settings, and read stats across your LinkedIn sender accounts. You sign in to HeyReach and pick the workspaces. Every tool says whether it reads, changes, overwrites or deletes data, so your assistant can ask you to confirm first. The tool list in its docs has no lead search tool, so the leads come from your lists or another source (HeyReach MCP connector).

Fits better than traxy if you run many LinkedIn sender accounts, for example as an agency, and already have your lead lists.

Open-source linkedin-mcp-server projects

The best-known open-source option is stickerdaniel/linkedin-mcp-server, with more than 4,000 GitHub stars when we checked. It runs on your machine and drives your own logged-in LinkedIn browser session, through a login window or a session imported from a browser you already use. Its tools read profiles, companies, jobs, posts, your feed and your inbox, search people and posts, and send connection requests and messages, with messages sent after confirmation. It describes itself as an independent project, not affiliated with LinkedIn or Microsoft. Smaller projects, such as eliasbiondo/linkedin-mcp-server, stick to reading and searching people, companies and jobs.

Fits better than traxy if you want to read and manage your own LinkedIn from Claude Desktop, Claude Code, Cursor or Codex with no vendor account, and you are happy to run the server yourself.

What does this look like in practice?

These are public workflows from the people who build traxy, run through the traxy MCP.

• A campaign and the inbox from chat. On November 9, Ben Buaron at traxy showed the flow from Claude, Codex or Grok: ask for an outreach campaign to everyone liking your LinkedIn content, then ask "who replied to me today?" The MCP pulls the LinkedIn and email inbox, prepares each reply, and sends it once you approve it in the chat (Ben's post).

• Social intent search as a search engine. Adam Salmanov, who builds traxy, connects the traxy MCP to Claude or ChatGPT and asks for people asking for what you sell, for example "find me founders in the US looking for help with LinkedIn outbound". He gets the people who liked, commented or posted about it. Other searches he suggests: "founders looking for aeo help", "agency owners hiring their first sdr" and "heads of revops unhappy with their crm" (Adam's post).

• A daily agent in Grok. Every morning Ben's agent in Grok asks traxy for this week's posts about outbound not working and posts sharing GTM breakdowns, tracks the best ones, pulls who liked and commented, and scores them against his ICP, so he opens the chat to a ready list of high intent leads (Ben's post).

• Competitor engagers into meetings. Ben's agency tracked competitors and agency owners selling to the same ICP, pulled the new ICP-matched engagers into Claude through the traxy MCP every day at 9am, based each message on traxy's signal and post context, and booked 3 meetings in a week (Ben's post).

• One prompt, 210 leads. Ben asked traxy in plain English to "find me leads that need outbound tools based on social intent" and got 210 leads back (Ben's post).

Which LinkedIn MCP server should you pick?

FAQ

What's the best LinkedIn MCP server?

traxy, if you want one server that finds high intent leads from social intent and runs the LinkedIn and email outreach to them: a 50%+ ICP match, the post or comment behind each lead, research on each one, and campaigns from your own accounts that send only after your yes. If you already have your list and only need sending, HeyReach, proReach, ReachHunt or LinkedNav fit better.

What's the best LinkedIn MCP server for Claude?

traxy, for LinkedIn GTM. You add it in Claude as a custom connector with sign-in and no API key, then find leads, research them, and draft, test and launch LinkedIn and email campaigns, with Claude waiting for your yes before anything sends. If you want Claude to draft each first LinkedIn message from public-data prospecting, BeReach fits better, and for local access with no vendor account, an open-source server does.

Which LinkedIn MCP server is best for posting or data extraction?

For posting, Composio's LinkedIn toolkit creates, comments on and deletes posts with OAuth sign-in. For data extraction, Apify runs LinkedIn scrapers from its store, and Bright Data runs pre-built LinkedIn scrapers and searches LinkedIn profile and company datasets. If you want the high intent leads and the LinkedIn and email outreach to them, traxy fits, and nothing sends until you say yes.

Which LinkedIn MCP servers find high intent leads, not just send messages?

traxy, LinkedNav and BeReach. traxy brings people who post, comment, like, view your profile or engage with the posts you track, keeps only those with a 50%+ ICP match and shows the activity behind each lead, then runs the campaign to them. HeyReach's MCP connector lists no lead search tool, and most other servers search LinkedIn or pull engagers and leave the fit check to your AI.

Which MCP servers let ChatGPT run LinkedIn outreach?

traxy, ReachHunt, HeyReach, Linked API, BeReach and Periodix document ChatGPT or OpenAI clients. With traxy, ChatGPT can find the high intent leads, run LinkedIn and email campaigns and answer replies, and it waits for your yes before anything sends. lemlist, La Growth Machine, Expandi, Waalaxy, Oppora, Salesbot and Reply.io document ChatGPT setup too.

Which LinkedIn MCP servers can answer replies from Claude?

traxy, HeyReach, lemlist, La Growth Machine, Salesbot, Reply.io and stickerdaniel/linkedin-mcp-server can send a reply from Claude, and each describes a way to confirm before it sends. Expandi drafts replies for you to approve, as Spurly does by default, and Oppora's MCP can't read reply text. With traxy, Claude reads the LinkedIn and email inbox, prepares each reply and waits for your yes before it sends.

Can Claude launch a LinkedIn and email campaign for me?

Yes, with traxy: Claude drafts the campaign, tests it on sample leads with nothing sent, shows who joins and who is skipped, and launches it after your yes, then it runs from your own LinkedIn and email accounts. lemlist, La Growth Machine, Oppora and Reply.io can also launch campaigns with email and LinkedIn steps from chat, HeyReach and Salesbot start LinkedIn campaigns, Waalaxy's MCP only enrolls prospects into existing campaigns, and Expandi leaves the launch to you.

Can the Salesforge MCP find leads, or only send to contacts I upload?

It sends to contacts you upload. Salesforge's MCP FAQ says it doesn't find new leads and points to Leadsforge, connected with its own API key; from chat it launches campaigns with email and LinkedIn steps and answers replies, and its page doesn't state whether the assistant asks before sending (checked November 2026). traxy brings the high intent leads and the LinkedIn and email campaign to them, with a yes in chat before anything sends.

Which LinkedIn MCP servers can read who engaged with a post?

traxy reads the engagers of any LinkedIn post you paste, yours or a competitor's, and keeps only the ones with a 50%+ ICP match before it runs the outreach. La Growth Machine, Oppora and Expandi also pull post engagers, Postiv reads comments only on posts in your own workspace, and vlpmedialtd/linkedin-mcp, stickerdaniel/linkedin-mcp-server, HeyReach and Waalaxy list no engager tool.

Does LinkedIn have an official MCP server?

Not for outreach. LinkedIn's Hiring Pro connector in Claude, announced October 21, 2026, covers Hiring Pro job posts only. The LinkedIn MCP servers for prospecting and outreach in this comparison, traxy included, are third-party and not affiliated with LinkedIn.

Which LinkedIn MCP servers send email as well as LinkedIn?

traxy does: email steps run in the same sequence as LinkedIn steps, from the mailboxes you connect, and replies from both land in one inbox. The other MCP pages we checked center on LinkedIn. BeReach and LinkedNav list email lookups, LinkedNav's agent guide shows a LinkedIn and email sequence in an example brief, and ReachHunt pairs LinkedIn with X instead of email.

Is it safe to connect a LinkedIn MCP server to my LinkedIn account?

No tool can promise your account won't be restricted, traxy included. LinkedIn's help page on prohibited software says it doesn't permit third-party software that automates activity on its website. You can still send from your own account, stay well inside LinkedIn's limits and keep every send behind your yes. Our LinkedIn API guide quotes LinkedIn's rules.

Do I need an API key to connect traxy?

No. The traxy MCP uses OAuth sign-in: you add https://api.traxy.ai/api/v1/mcp to your client, sign in to traxy, pick your workspaces and approve. traxy doesn't accept a REST API key as an MCP credential. Among the others, proReach and Periodix use API keys, LinkedNav asks for an API key on its connect page, Linked API puts tokens in the connector URL, and ReachHunt, HeyReach and BeReach use sign-in.

Does every traxy customer get the MCP?

Yes. MCP access is included on every traxy plan, campaigns, sequences, senders and inbox replies included. Plans are on the pricing page.

Can I use traxy with another MCP server in the same chat?

Yes. Claude and ChatGPT can call several servers in one conversation, for example traxy for the leads and outreach and a CRM server to check who is already a customer. Keep one tool sending to each lead, so no one gets two sequences. From chat, traxy can also send leads to a Webhook, Clay, n8n or Zapier destination.

How do I run LinkedIn outbound from Claude Code or Cursor?

With traxy, open Integrations, pick Claude or Cursor and follow its setup to add https://api.traxy.ai/api/v1/mcp with OAuth sign-in; a REST API key isn't accepted. Then ask for the leads, draft and test the campaign, and say yes to launch it from your own LinkedIn and email accounts. PumpGTM, FirstTouch, LinkMCP, SyncGTM, Datamagnet and Manifold also document a claude mcp add command on their own pages.

Which MCP servers find the people engaging with a LinkedIn post?

SyncGTM, Unipile, Crispy and Apify actors return a post's reactors or commenters, but their pages describe no ICP check, and SyncGTM needs a separate tool for emails. traxy keeps only the engagers with a 50%+ ICP match, shows the post or comment behind each lead, and runs LinkedIn and email outreach to them in the app, through trax or from the MCP after your yes in chat.

Is an MCP server's approval step the same as LinkedIn's permission?

No. A yes in the chat or a review queue is the tool's own control over what you send, not authorization from LinkedIn. LinkedIn's help page says it doesn't permit third-party software that automates activity on its website, and no tool can promise your account won't be restricted, traxy included.

Related reading

• The traxy LinkedIn MCP

• What the traxy MCP server can do, by job

• Is there a LinkedIn API for outreach?

• Claude MCP servers for sales teams

• How to find and research LinkedIn leads from Claude

• The best LinkedIn GTM tools

• What is social intent search?

• LinkedIn outbound with traxy