The LinkedIn GTM MCP Workflow: Find, Research, Launch and Reply From Claude or ChatGPT

TL;DR: To run your whole LinkedIn GTM from Claude or ChatGPT, connect the traxy MCP (endpoint https://api.traxy.ai/api/v1/mcp, a one-time browser sign-in, no API key) and ask in plain English. 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. From one chat you set your ICP and Tracking, find and research leads, build, test and launch a campaign, and answer the replies. Your AI shows you what it will do and waits for your yes before anything launches or sends, and once you approve a campaign it runs on its own, or you approve every message.

Checked November 12, 2026. traxy wrote this guide and makes the traxy MCP. The steps come from the traxy docs, linked where they appear. Which tools your AI sees depends on your grant, role, workspace and plan, so the list your client returns is the reference for your connection.

How do I run my whole LinkedIn GTM from Claude or ChatGPT?

Connect traxy to Claude or ChatGPT as an MCP server, then ask your AI to do each part of the work: traxy finds high intent leads from LinkedIn social intent, researches each one and runs LinkedIn and email campaigns from your own connected accounts. Your AI asks for your yes before anything launches or sends.

The whole workflow is five steps, and each one is explained further down with the prompts to use:

1. Connect traxy and pick workspaces. Add the traxy MCP to your AI client, sign in to traxy once and choose which workspaces it can use.

2. Set your ICP, signals and tracking. Ask your AI to show your ideal customer profile (ICP) and Signal Brief, propose changes and apply them after you confirm. Add the people, posts and company pages to track.

3. Find high intent leads and research them. Search 71M+ scanned LinkedIn likes, posts and comments in plain English, read your qualified leads with the post or comment behind each one, and have traxy research the ones you pick.

4. Build, test and launch a LinkedIn and email campaign. traxy prepares the campaign and sequence, tests it on sample leads with nothing sent, shows who joins and who is skipped, and launches when you say yes.

5. Answer replies and check results. Ask who replied, approve the answers traxy prepares, and read campaign and lead results in the same chat.

It works the same way from Claude, ChatGPT, Codex, Cursor and Grok. You can stop at any step and finish in the traxy app, or hand a step to trax, the agent inside traxy.

What is the LinkedIn GTM MCP?

The LinkedIn GTM MCP is the traxy MCP server: one connection that lets an AI client run traxy, covering both halves of LinkedIn GTM. The first half is finding high intent leads from social intent, meaning what people post, comment and like on LinkedIn, who views your profile, and who engages with competitors', creators' and company pages' posts. The second half is the outreach to those people over LinkedIn and email. How the whole workflow runs, and how it differs from sales engagement, data and LinkedIn automation tools, is explained in what is LinkedIn GTM.

MCP (Model Context Protocol) is a standard way for an AI assistant to use another product's tools. traxy publishes a remote MCP endpoint at https://api.traxy.ai/api/v1/mcp. Your client connects with OAuth: you sign in to traxy in the browser, choose the workspaces and review the actions you allow. A traxy REST API key is not accepted as an MCP credential (connect an AI client).

What makes it a GTM MCP rather than a messaging MCP is what traxy brings to the chat. Most LinkedIn MCP servers send messages or pull profiles. traxy brings the leads (each one a 50%+ ICP match, with the post or comment behind it), the reason to reach out and the campaign. The public tool catalog lists 234 tools before scope and workspace filtering, grouped by job: leads, discovery and search, targeting, Tracking, campaigns, sequences and senders, research and drafts, inbox, analytics and integrations (traxy MCP tool list).

How it fits with the rest of traxy:

• The traxy app has a section for each step: Leads, lead discovery, Tracking, Outbound (campaigns, sequences, senders) and the inbox.

• trax is the chat inside the app. It can search for leads, set your ICP and signals, research and build campaigns, with an approval card before anything changes your setup.

• The traxy MCP brings what trax does into your own AI client, so the LinkedIn work sits next to the rest of your day in Claude or ChatGPT.

• The REST API reads leads, signals and engagements from your own code. It is separate from the MCP.

Here is the workflow in one table, with what you ask and what waits for you:

Step 1: connect traxy and pick workspaces

Open Integrations in traxy, pick your AI client and follow the steps shown there. It takes a few minutes, and you only sign in once.

1. Pick your client in traxy. Integrations lists Claude, ChatGPT and Grok. The ChatGPT entry opens the Codex connection guide, which covers the ChatGPT desktop app and Codex (connect ChatGPT or Codex). For another compatible client, use its remote MCP setup with the endpoint https://api.traxy.ai/api/v1/mcp.

2. Add the server in your client. In Claude, add the URL as a remote custom connector (connect Claude); on a work account, your admin may need to allow custom connectors first. In Claude Code, run claude mcp add --transport http traxy https://api.traxy.ai/api/v1/mcp, then type /mcp to sign in. In Codex, run codex mcp add traxy --url https://api.traxy.ai/api/v1/mcp. In Cursor, add a traxy entry with the same URL to .cursor/mcp.json.

3. Sign in and choose workspaces. Complete the browser sign-in, choose a default workspace and any others, and review the permitted actions for each one before you approve. One connection can authorize up to 50 workspaces you select (MCP workspaces).

4. Check the connection. Ask your AI to list your authorized workspaces, then show five qualified leads from the one you pick. Review that result before asking for anything that changes data or starts discovery.

Two details save confusion later. An existing connection does not gain new workspaces or permissions on its own, so authorize again when you need more access. And there is no "current workspace" on the server side: if you connected several, name the workspace in your request, or calls go to the default. That matters for agencies running traxy for several clients and for teams with more than one workspace.

Prompts to try:

• “List my authorized traxy workspaces, then show five qualified leads from the Acme workspace.”

• “Check my traxy MCP connection status and tell me which workspace is the default.”

Full client-by-client setup lives in two guides: how to connect LinkedIn to ChatGPT and finding LinkedIn leads in Claude with traxy.

Step 2: set your ICP, signals and tracking from chat

Ask your AI to read your current ICP and Signal Brief, suggest changes, and apply them once you confirm. These two settings decide who counts as a lead, so set them before you search or launch anything.

When you start traxy from your website, it drafts your ICP and the signals to watch, and you fix whatever is off. From chat you can then read and update the ICP, including exclusions, and the Signal Brief, with revision history for both. The docs ask the AI to read the current version before it updates it, and fields you leave out stay as they are (traxy MCP tool list).

What you can set from chat:

• ICP. The roles, seniority, locations, company sizes and industries that fit, plus who to exclude. Only people with a 50%+ match to your ICP become leads.

• Signal Brief. The topics and evidence that make a lead relevant to what you sell (intent signals).

• Tracking. Follow people (competitors, influencers, teammates, investors), single posts by URL and company pages by domain. traxy qualifies whoever engages with them (Tracking). You can list, add, update, pause, remove or restore tracked profiles and tag them. There are no tracking limits on profiles, posts or company pages.

• Blocklists. Companies and people you never want as leads, with a preview before the change.

Profile views are a signal too. In the app, Targeting > Signals > LinkedIn activity has a Profile views switch, and traxy picks up profile views of the sender accounts you connect. Tracking a company page finds people who engage with that page's posts; it doesn't list every employee at the company.

Prompts to try:

• “Show my active ICP and Signal Brief. Suggest three changes so we focus on B2B SaaS founders with 10 to 50 people, and wait for my OK before you update anything.”

• “Track this LinkedIn profile and tag it as a competitor: {profile URL}.”

• “Add this post to Tracking so traxy qualifies everyone who engages with it: {post URL}.”

Step 3: find high intent leads and research them

Ask your AI for the people you want and the need they show, and traxy returns people who fit your ICP with the LinkedIn post, comment or like behind each one. Then pick the best of them and have traxy research each lead before any outreach is prepared.

Four ways to find leads from chat:

• Social Intent Search. Describe the person and the need in plain English, for example “founders in the US looking for help with LinkedIn outbound”, and traxy searches 71M+ scanned LinkedIn likes, posts and comments. Each person comes with the activity behind them. Adam Salmanov at traxy runs this exact prompt through the traxy MCP (his post).

• Your qualified leads. List and filter the leads traxy already found for your workspace, labeled Warm, Related or Activity, and read lead counts and stats.

• A discovery run. Ask traxy to look for new people who fit your ICP and signals. Your AI asks you to confirm first, because new qualified leads count toward your capacity.

• One post's audience. Paste a LinkedIn post URL and ask traxy to qualify the people who engaged with it. This also waits for your confirmation.

From the same chat you can find people at companies hiring for the roles you name, and ask for the leads who viewed your profile once Profile views is on.

Research and enrichment. For any lead, your AI can read its signals, recorded engagements and the posts behind them, plus account reports. Hand traxy up to 25 leads at a time and it researches each one in the background and prepares a draft from what it finds. Company research covers Meta ads and website traffic. For contact data, request an email or phone lookup, or turn on automatic email enrichment for new leads. Signals outside LinkedIn, such as hiring, funding or a job change, come from research on each lead, not from a live event feed.

Sort before you send. Mark leads good, bad, done, snoozed or client, add tags and notes, then save the people you want to reach as a shared lead view. Campaigns use shared views as their audience, so this view is the bridge into step 4.

Prompts to try:

• “Find me founders in the US looking for help with LinkedIn outbound, with the post or comment behind each one.”

• “Show my Warm leads from the last 7 days who work at software companies with 10 to 200 people.”

• “Research these 10 leads: what their company sells, how they sell today and what they posted about recently. Then save the ones who fit as a shared view called “Outbound founders”.”

Step 4: build, test and launch a LinkedIn and email campaign

Ask your AI to build a campaign on your shared view, test it on sample leads, and show you the launch preview. Nothing sends during the test, and the campaign launches only after you say yes in the chat.

This is the outreach half of traxy, and it runs from the same chat. For how it works in the app, see LinkedIn outbound with traxy.

1. Check your senders. Messages go out from your own connected LinkedIn accounts and mailboxes. Your AI can list your senders and open the connect flow for a new one. A sequence with LinkedIn and email steps needs a sender name with both channels connected (sender accounts). Each sender's limits, active hours and ramp-up still apply, and a campaign doesn't bypass them.

2. Draft the campaign. Name the audience (your shared view), the research facts you want on each lead and your offer. traxy researches every lead for those facts, filters out people who don't fit and groups the rest by signal, with a personalized message variant per group. When a required research fact isn't found, the lead stays in traxy, and you can choose to skip it in the campaign.

3. Set the sequence. For example, a LinkedIn connection request, a message after they accept, then an email follow-up a few days later. You can use a saved sequence template or have traxy prepare one (sequences and one-off messages).

4. Test on sample leads. A test prepares the research and message previews without sending anything. Read whether the research supports each line, and look at skipped leads as well as good drafts (review and launch).

5. Review the launch preview. It shows who joins, who is excluded because another unfinished campaign holds them, people a completed campaign already contacted, the sender, the steps and the delivery mode.

6. Choose the delivery mode and launch. Approve each message in the campaign's review queue, or let the campaign send on its own. Your yes to the launch is the campaign launch approval.

That last choice is what "on autopilot" means in traxy: you test the campaign and approve the launch once, then it runs on its own, or you keep the review queue and approve every message. Switching a campaign to automatic later doesn't release messages already waiting for approval; review those separately.

Run many campaigns at once. You can run a campaign per signal, audience or offer, for example competitor engagers, profile viewers and people posting about your problem. Campaigns run in priority order, and a lead goes to the first campaign it fits and isn't researched or messaged by two unfinished campaigns at once (manage campaigns). Ask your AI to reorder campaigns when audiences overlap.

Prompts to try:

• “Build a LinkedIn and email campaign on my shared view “Outbound founders”. Research each lead’s recent LinkedIn activity, group them by signal and open with what they engaged with, no pitch. Test it on 3 leads first.”

• “Show me the launch preview: who joins, who gets skipped and why, which sender and which steps.”

• “Launch it with every message waiting for my approval in the review queue.”

traxy co-founder Ben Buaron asks his AI chat for an outreach campaign to all the leads liking his LinkedIn content (his post).

Step 5: answer replies and check results

Ask your AI who replied, and through the traxy MCP it reads your LinkedIn and email conversations from the traxy inbox, prepares an answer for each and sends it from your own account once you approve it in the chat.

The traxy inbox keeps one conversation per lead across LinkedIn and email (Inbox). An incoming reply stops active sequence outreach for that lead, and so do unsubscribes and hard bounces. From chat you can also mark conversations read or unread, archive them, pause, resume or stop a lead's sequence, unsubscribe a lead and add people to your do-not-contact list. A reply you send from the inbox is a separate action from approving a campaign's queued message.

Ben Buaron asks his AI chat “who replied to me today?”, and traxy pulls the LinkedIn and email inbox, prepares each reply and sends it once he approves it in the chat (his post).

Check results in the same chat. Your AI can read campaign performance, the stage of each lead in a campaign, aggregate analytics, post analytics with the qualified leads each post brought in, and attributed meetings when Google Calendar is connected. You can also set a daily brief. Leads can go on to other tools by webhook, Clay, n8n or Zapier, and to Slack or HubSpot through those integrations.

Prompts to try:

• “Who replied to my campaigns today? Sort them by interest and prepare an answer for each. Wait for my OK before sending.”

• “Pause the sequence for this lead; I will take the conversation from here.”

• “How did my campaigns do this week? Which posts brought in the most qualified leads?”

What waits for your yes, and what the MCP can't do

Anything that launches a campaign, sends a message, replies in the inbox, starts discovery that uses capacity or changes your setup waits for your yes in the chat. Your AI shows a preview first, and a draft or preview never counts as approval.

What the MCP can't do, or won't do on its own:

• It doesn't log in to LinkedIn. The sign-in connects your AI to your traxy workspace. Sending needs sender accounts connected in traxy.

• It can't go beyond your grant. Your role, workspace access and plan decide which tools work. Campaign actions depend on your workspace's access and plan; a tool in the catalog isn't a promise that it works in every workspace.

• It can't recall a sent message. Pausing a campaign stops further work, not messages already delivered.

• It doesn't skip sender limits. Approval lets the next step proceed, but delivery can still wait for sender availability, schedule and limits. A status of "May have been sent" shouldn't be retried blindly.

• It doesn't watch the whole web. Lead finding is LinkedIn social intent. Hiring, funding and other outside signals come from research on each lead.

• It doesn't promise results. A like or comment shows interest in a topic, and the match score measures fit, not the chance that someone buys. No tool can promise meetings, revenue or zero LinkedIn restriction risk.

When a different setup fits better: if you already have your list and only need your AI to send messages, a send-only LinkedIn MCP server can be enough; the LinkedIn MCP servers compared covers them from their own pages. If you'd rather click through each step, the same workflow runs by hand in the traxy app or through trax. And if your company's policy rules out AI clients, leads can reach your team by Slack, webhook or CSV export instead.

For every MCP job with its tool names, read what the traxy MCP server can do. For 25 ready-to-copy prompts grouped by job, see Claude and ChatGPT prompts for LinkedIn GTM. For the product end to end, read how traxy works.

FAQ

Can I run my whole LinkedIn GTM from ChatGPT, not only Claude?

Yes. The traxy MCP works the same from ChatGPT, Claude, Codex, Cursor and Grok. In traxy, the ChatGPT entry under Integrations opens the Codex connection guide for the ChatGPT desktop app and Codex.

Does my AI send LinkedIn messages without asking me?

No. Your AI shows what will launch or send and waits for your yes in the chat. After you approve a campaign launch, it runs on its own, or you choose the review queue and approve every message.

Do I need a traxy API key to connect the MCP?

No. The MCP uses a browser OAuth sign-in. A traxy REST API key is not accepted as an MCP credential, so leave token fields empty.

Who writes the outreach messages?

traxy prepares the drafts from the research on each lead and the signal behind it, and you review them in the test and, if you choose, in the review queue. You can edit or rewrite any draft before it is approved.

Can I run several campaigns at once from chat?

Yes. You can run a campaign per signal, audience or offer. Campaigns run in priority order, and a lead isn't messaged by two unfinished campaigns at once.

Can an agency run this for several clients?

Yes. An agency can run traxy for its clients, and one MCP connection can authorize up to 50 workspaces you select. Name the workspace in each request, since calls without one go to the default.

Is the traxy MCP the same as trax?

No. trax is the chat inside the traxy app. The traxy MCP brings the same work, including campaigns, senders and the inbox, into your own AI client.

Which LinkedIn signals can I start from?

Posts, comments and likes about the problem you solve, profile views, people who engage with tracked profiles, posts and company pages, and Social Intent Search across 71M+ scanned LinkedIn likes, posts and comments.

Related reading

• LinkedIn GTM MCP: run traxy from your AI chat

• LinkedIn and email outreach with traxy

• What the traxy MCP server can do

• How to connect LinkedIn to ChatGPT

• Find LinkedIn leads in Claude with traxy

• The LinkedIn MCP servers compared

• How traxy works