How to send a LinkedIn message, connection request or email to one lead from Claude or ChatGPT

TL;DR: Yes. Connect traxy to Claude or ChatGPT through the traxy MCP, pick a lead, and ask for a LinkedIn message, a connection request or an email. traxy prepares the draft from the post or comment that made that person a lead, shows you the message, the sender account and the channel, and sends it from your own connected account only after you say yes in the chat. The same chat then reads the reply and answers it, and it works the same way from Codex, Cursor and Grok.

This guide covers the whole one-lead workflow: what you need first, where the lead and its reason come from, the prompts that work, how the preview and your yes work, when a one-off message is the right tool and when the lead belongs in an approved campaign, how to follow up on the reply, and the limits. Product details come from the traxy MCP tool catalog, the sequences docs and the sender accounts docs, checked November 2026.

Can Claude or ChatGPT send a LinkedIn message, connection request or email to one lead for me?

Yes, when the AI client is connected to a tool that can send from your accounts. With the traxy MCP, Claude or ChatGPT asks traxy to prepare a one-off message to the lead you picked, shows you the preview, and sends it from your connected LinkedIn account or mailbox after you approve that exact message in the chat.

On its own, an AI chat cannot reach LinkedIn or your mailbox. It can write text, and then you copy it into LinkedIn and press send yourself. That works for one message, but the AI does not know who the person is, why you are writing to them now, or whether someone on your team already messaged them. There are three common setups:

1. Copy and paste. You describe the person to Claude or ChatGPT, it suggests a message, and you send it by hand on LinkedIn or from your email. The AI only knows what you tell it.
2. A LinkedIn MCP server that sends. Some MCP servers let your AI send messages or pull profiles for a list you already have. What it can send, from which account, and whether it asks first depend on that tool. If you already have your list and only want your AI to press send, the tools in the LinkedIn MCP servers compared may fit better.
3. traxy. The lead comes with its reason: the LinkedIn post, comment, like or repost that put them on your list. traxy prepares the message from that signal, you can check the lead's outreach status first, and it sends from your own connected sender account after your yes.

We make traxy, so weigh that. The rest of this guide covers the third setup.

In the traxy app the same action is called a one-off message: open a lead, choose Message, check To, From and the channel, then Send. It covers a single LinkedIn message, connection request or email to one lead. From chat, the traxy MCP has its own tools for it, listed in the tool catalog:

What does it need: the traxy MCP connected, and a lead with its signal?

Three things: the traxy MCP connected to your AI client, a connected sender account to send from, and a lead in your traxy workspace, which comes with the signal behind it.

1. Connect the traxy MCP once

1. Open Integrations in traxy and pick your client: Claude, ChatGPT (which opens the Codex connection guide) or Grok. Other clients that support remote MCP with OAuth, such as Cursor, use the server URL https://api.traxy.ai/api/v1/mcp.
2. Add it in your client. In Claude you add the URL as a remote custom connector and start the connection (Claude setup).
3. Sign in and pick workspaces. The connection uses OAuth. You choose a default workspace and any others, and review what each workspace allows before you approve. No traxy REST API key is needed, and a REST API key is not accepted as an MCP credential (connect an AI client).
4. Check it. Ask your AI to list your workspaces and show five qualified leads from the one you will use.

If you already find leads or answer replies from chat, you are connected and can skip this step. The same connection runs the rest of your LinkedIn GTM from chat.

2. Connect a sender account

Messages go out from LinkedIn accounts and mailboxes you connect under Outbound > Sender accounts, or you can ask trax, the in-app traxy agent, to help you connect one. Accounts are grouped by sender name, and each account keeps its own limits, active hours and ramp-up settings (sender accounts docs). A LinkedIn message or connection request needs a connected LinkedIn account; an email needs a connected mailbox and the lead's email address. A LinkedIn and email sequence needs a sender name with both channels connected.

3. A lead with its signal

This is the half of traxy that comes before any message. 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 means you test a campaign on sample leads, review the launch and approve it once, and then it runs from your connected accounts. Or you approve every message. A one-off message from chat is the single-lead version of that outreach, and it always waits for your yes.

The leads come from social intent:

• Post engagement. People who write, comment on or like posts about the problem you solve.
• Tracking. People who engage with the competitors, creators, single posts or company pages you add to Tracking.
• Profile views. People who view the LinkedIn profiles of the sender accounts you connect.
• Social Intent Search. A plain-English search over 71M+ scanned LinkedIn likes, posts and comments, from the app or from chat. See Social Intent Search.

A person counts as a lead only with a 50%+ match to your ideal customer profile, and every lead shows the post, comment, like or repost that put them on your list. That source is what makes a one-off message worth sending: you write to someone about the thing they said or reacted to, not about your product in general. For finding and researching those leads from chat, read how to find and research LinkedIn leads from Claude with traxy.

How does it work: you ask in chat, and traxy prepares the draft from the lead's signal and research?

You name the lead and the channel, the AI reads the lead's signals and research through the traxy MCP, traxy prepares a draft, and you edit it in the chat. Nothing sends at this stage.

A good first prompt:

"In my traxy workspace, open the lead who commented on the post about outbound reply rates this week. Show me their comment, their role and company, and whether anyone on my team has contacted them. Then prepare a short LinkedIn connection request note that refers to their comment. Show me the message, the sender account and the channel. Don't send anything until I say yes."

Behind that prompt, step by step:

1. Find the lead. The AI lists and filters your qualified leads, or searches social intent if the person is not in your workspace yet. Starting a discovery run or qualifying a post's audience to add new leads asks you to confirm first.
2. Read the signal. The lead tools return the lead's signals, recorded engagements and the posts behind them. This is where the message gets its reason: the exact comment, its date and the post it was on. Check the date. A comment from last week is a better opener than one from months ago.
3. Read the research. If the lead was researched in a campaign, the facts it found are there to use. You can also ask the AI to look at the lead's company or role before you write. For an email, ask for a contact lookup first if the lead has no email address yet; lookups need an explicit request.
4. Check the outreach state. Ask for the lead's outreach status so you know whether they are already in an active sequence or were messaged before. A lead can be in only one active sequence at a time.
5. Prepare the draft. traxy prepares the message with traxy_draft_outreach_message. You read it in the chat and ask for changes: shorter, a different question, no link, your tone from an earlier message. Each change is a new draft.

Other prompts that work:

• "Show me my three best-fit leads who viewed my profile in the last 7 days. For the first one, prepare two short connection request notes and let me pick one."
• "This lead liked two posts by a competitor I track. Prepare an email that mentions the topic of those posts, not the competitor. Use my main mailbox."
• "Prepare a LinkedIn message to this lead in reply to their post about hiring their first SDR. Two sentences, one question, no pitch."
• "Which of my senders can message this lead on LinkedIn? List them first."

A few habits keep a one-off message good. Refer to the signal in the first line, in plain words, so the person knows why you are writing. Ask one question. Keep a connection request note short. And write as the person whose account sends it, since the lead sees that name.

The yes in chat: does anything send before you confirm, and which account sends it?

No. The AI shows you the preview, you approve that exact message in the chat, and only then does it go out, from the connected LinkedIn account or mailbox shown in the preview.

1. Preview. The AI calls traxy_preview_outreach_message and shows you the final text, To (the lead), From (the sender account) and the channel: LinkedIn message, connection request or email. This mirrors the app, where you check To, From and the channel before you press Send.
2. Your yes. You reply in the chat with a clear yes for that message. The tool catalog says to preview message sending before confirming, and that a draft or preview does not authorize execution. If you change a word after the preview, ask the AI to show it again before you approve.
3. Send from your account. traxy_send_outreach_message sends the one-off message from the sender you approved. For a single step, Send is the approval (sequences docs).
4. Check the result. Ask for the status with traxy_get_outreach_status.

The status tells you what happened, and the four values mean different things:

Two details for teams. Your connection only has the tools your workspace role, product access and scopes allow; seeing a tool in the catalog does not grant permission to use it in every workspace, and tools/list shows what your connection can use. If your client does not offer the send tool, ask your workspace admin about your access. And if you connect several workspaces, for example one per client at an agency, name the workspace in each request so the message goes out from that client's sender accounts.

Should you send a one-off message or add the lead to an approved campaign?

Send a one-off message when you picked one person by hand and want to write to them once. Put the lead in a campaign when you want follow-ups, or when the same reason applies to many people.

How each path works from chat:

• Adding follow-ups turns a message into a sequence. In traxy, adding follow-up steps to a lead's message creates a sequence, and activating it approves the activation and enrollment for that lead (sequences docs). Do this when the person matters enough for a second touch but does not fit a running campaign.
• A campaign for many people. Campaigns run on shared lead views. Ask the AI to save a view of the leads you want, filter-based or a fixed list, then draft the campaign, test it on sample leads and preview the launch. The launch review shows who joins, who is skipped because another unfinished campaign holds them or a completed campaign contacted them before, the sender, the steps and the delivery mode (review and launch). You choose whether you approve each message or the messages send automatically after the launch (create a campaign). Launching waits for your yes in the chat.
• A lead already in a campaign. Because a lead can be in only one active sequence, check the lead's outreach status before you send anything by hand, so your one-off message does not land between two campaign steps.

A common pattern is to start with one-off messages and then turn what works into a campaign. You write to three people who commented on the same post, see which opener gets answers, and then ask the AI to build a campaign for everyone else who engaged with that post, with that opener as the message for that group. You can run many campaigns at once on different signals. For a full run from connection to launch and replies, with a prompt for each step, read the LinkedIn GTM MCP workflow, and for how outreach works in the app, read LinkedIn outbound with traxy.

How do you follow up on the reply from the same chat?

Ask who replied. The AI reads the conversation from the traxy inbox, traxy prepares an answer, and it sends from the same connected account after your yes.

Every reply lands in Outbound > Inbox, where each lead has one conversation across LinkedIn and email. From chat:

1. List conversations with traxy_list_inbox_conversations, for example only unread ones, or only this lead's.
2. Read the thread with traxy_get_inbox_conversation: the messages on both channels, the sender accounts, any sequence enrollment and the do-not-contact state.
3. Prepare and approve the answer. traxy_reply_inbox_conversation takes confirmSend: true, which the client should pass only after you approve the exact reply.
4. Act on what they said. Tag the lead for a teammate, pause or stop a sequence, or, if they asked you to stop, unsubscribe them, which adds them to Do not contact. Ask the AI to show you what will change before it applies any of these.

A prompt for the next morning: "Did the lead I sent a connection request to yesterday accept or reply? If they replied, show me the thread and prepare a short answer. If they accepted and haven't replied, prepare a first LinkedIn message about their comment, and don't send it until I say yes."

If the lead was in a sequence, their reply has already stopped it on both channels, so you do not race your own follow-up. Every step of reply handling from chat, including pausing a single lead and unsubscribes, is in how to answer LinkedIn outreach replies from Claude.

What are the limits?

• Every send waits for you. A one-off message from chat goes out only after your yes on that message. traxy is not for teams that want messages to go out with no review at all.
• Sending needs connected sender accounts. A LinkedIn message or connection request needs a connected LinkedIn account; an email needs a connected mailbox and the lead's email address.
• LinkedIn's own rules still apply. LinkedIn generally lets you message people you are connected with, so a connection request is often the first touch. Whether a request is accepted, or a message is read, is up to the person; traxy does not guarantee delivery or acceptance.
• Timing follows your senders. Limits, active hours and spacing on a sender account can delay a send, and the chat does not bypass them.
• Your access decides your tools. Workspace roles, product access, scopes and each action's checks apply. A tool you see in the catalog may not be available in every workspace; tools/list shows what your connection can use.
• One active sequence per lead. Check a lead's outreach status before sending by hand if they may be in a campaign.
• A sent message is sent. Pausing or stopping stops what comes next, not what already went out. And never resend a message marked May have been sent before you check the conversation.
• The reason is LinkedIn activity. The signal behind each lead is a LinkedIn post, comment, like, repost or profile view. Signals outside LinkedIn, such as hiring, funding or a job change, come from research on each lead, not from a live event feed.

FAQ

Can Claude or ChatGPT send a LinkedIn message to a lead for me?

Yes, with the traxy MCP connected. traxy prepares the message from the lead's signal, the AI shows you the preview with the sender and channel, and it sends from your own connected LinkedIn account after you approve it in the chat.

Can I send a LinkedIn connection request from Claude?

Yes. A one-off message in traxy can be a LinkedIn message, a connection request or an email to one lead. Ask for a connection request, check the note, the sender and the channel in the preview, and say yes.

Will Claude or ChatGPT send anything without asking me?

It should not. The traxy MCP is built to preview message sending before you confirm, and a draft or preview does not authorize a send. Approve one message at a time, and ask to see it again if you change anything.

Which account does the message come from?

From a LinkedIn account or mailbox you connected under Outbound > Sender accounts. The preview shows the From account before you approve, and you can ask the AI to list your senders first.

Does it work from ChatGPT, Codex, Cursor or Grok?

Yes. The traxy MCP works the same way from Claude, ChatGPT, Codex, Cursor and Grok, and other clients that support remote MCP with OAuth can connect with the server URL. Every client waits for your yes before anything sends.

Do I need an API key?

No. The traxy MCP uses an OAuth sign-in where you pick the workspaces to grant, and a traxy REST API key is not accepted as an MCP credential.

What if the status says May have been sent?

Don't send it again. Read the conversation first. An uncertain result is not a failed send, and repeating it blindly can message the same person twice.

Should I use a one-off message or a campaign?

A one-off message suits one person you picked and one touch. A campaign suits many people with the same reason, follow-ups across LinkedIn and email, research on each lead and a launch you approve once, or every message approved in a review queue.

Related reading

• The LinkedIn GTM MCP and LinkedIn outbound with traxy
• How to find and research LinkedIn leads from Claude with traxy
• How to answer LinkedIn outreach replies from Claude
• What the traxy MCP server can do, by job
• The LinkedIn GTM MCP workflow, from connection to replies