Lead generation

How to run a daily LinkedIn lead search with an AI agent and the traxy MCP

Have an AI agent find this week's LinkedIn posts on your topic, qualify the engagers against your ICP through the traxy MCP, and start each day with leads.

By the traxy teamPublished Nov 12, 202615 min read

TL;DR: Yes. Connect traxy to an AI agent through the traxy MCP and give it one prompt to run each morning: find this week's LinkedIn posts on a topic your buyers care about, pick the best ones, track them and qualify the people who engage against your ICP. traxy keeps only people with a 50%+ ICP match as leads and shows the post or comment behind each one. The agent asks for your yes before it uses capacity on new leads, and again before any campaign launches or any message sends. The schedule itself lives in your AI client, not in traxy.

Checked November 12, 2026, against the traxy MCP tools reference. traxy wrote this guide.

This guide is for founders and GTM teams who already run recurring AI agents and want one of them to bring in fresh high intent leads every day. It covers what the agent does, how to connect it, the prompt to schedule, how to review the leads and how to turn them into LinkedIn and email outreach. If you want a one-off lead search from Claude instead of a daily routine, start with how to find and research LinkedIn leads from Claude.

Can an AI agent find new LinkedIn leads for me every morning?

Yes. Through the traxy MCP, an AI agent can find this week's LinkedIn posts on your topic, track the best ones and qualify their engagers against your ICP after you confirm the capacity use, so you start the day with high intent leads you can put into a campaign you approve.

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 here means the campaign runs on its own after you approve it, or you approve every message. The daily agent covers the first half of that work, finding the people, and hands the second half to a campaign.

Two things are worth knowing before you set it up:

The schedule belongs to your AI client. traxy gives the agent the tools. Running the prompt every morning is a feature of the client or agent platform you use, if it offers scheduled or recurring tasks. Where it doesn't, you run the same prompt yourself when you sit down.
The yes belongs to you. Qualifying a post's audience and starting a discovery run both need your agreement that new qualified people may use your workspace's capacity. Launching a campaign or sending a message needs your yes in the chat too.

One real example from traxy use cases: every morning Ben's agent in Grok Bot asks traxy for this week's posts about outbound not working and GTM breakdowns, then pulls and scores the people who engaged.

What the agent does each day: find posts, pick the best, qualify the engagers

The daily routine has four moves. Each one maps to a documented tool in the traxy MCP.

Find posts on your topic. traxy_find_posts finds LinkedIn posts for a topic. The agent asks for posts about the problem you solve, the role you sell to or a theme your buyers argue about, and keeps the recent ones.
Pick the best ones. The agent reads the candidates (traxy_read_post) and keeps the posts whose topic and commenters look closest to your buyers. A post about "our outbound stopped working" from a founder in your market beats a viral motivational post with a big audience.
Track the winners. traxy_track_posts adds single posts to Tracking, so they show up under Single posts with their qualified leads (Tracking).
Qualify the engagers. traxy_qualify_linkedin_post requests qualification for a post URL. traxy checks the people who engaged against your ICP and keeps only those with a 50%+ match as leads (how qualification works). This step needs your confirmation first, because new qualified people use capacity.

At the end, traxy_list_leads lists the new leads with the filters you ask for, and the agent writes you a short summary: the posts it used, who qualified, their ICP match and the comment or reaction behind each person.

There is a second route for days when you'd rather search by audience than by post. traxy_start_agent_run starts a confirmed discovery run for new people who fit your ICP and signals, and the same capacity confirmation applies. Posts suit timely topics, and discovery runs suit steady coverage of your ICP. For every MCP tool grouped by job, see what the traxy MCP server can do.

Connecting the agent to traxy

The traxy MCP works with Claude, ChatGPT, Codex, Cursor, Grok and other MCP clients. You add one URL and sign in to traxy once. No API key is needed (connect an AI client).

https://api.traxy.ai/api/v1/mcp

In traxy, open Integrations and choose Claude, ChatGPT or Grok; for Codex, Cursor or another client, add the MCP URL in that client.
Add the MCP URL in your client's connector settings.
Sign in to traxy, pick the workspaces the agent may use and approve the connection.
Run a read first: "List my authorized traxy workspaces, then show five qualified leads from the workspace I select."

Two setup details matter more for a daily agent than for a one-off chat:

Name the workspace in the prompt. Each call uses the workspace you name or the connection's default, not the one you last opened in the app.
Check the connection stays signed in. The MCP uses a sign-in, not a key. Before you rely on a scheduled run, test that your client still reaches traxy on a day you didn't open it. traxy_get_mcp_connection_status shows the connection state. If your setup needs a key-based connection, the traxy REST API reads leads, signals, tracked profiles, analytics and meetings from your own code, but finding and qualifying new leads runs through the MCP or the app.

Set your ICP and Signal Brief before the first run. The agent qualifies against whatever ICP is active, so a loose ICP gives you a loose list. You can read and update both from chat with a confirmation, or in the app under Targeting.

The prompt to schedule

Keep the scheduled prompt short and explicit about where the agent must stop. This version finds and ranks posts without asking, then waits for you before it uses capacity:

"In my {workspace} traxy workspace: find LinkedIn posts from the last 7 days about {topic 1} and {topic 2}. Skip posts older than a week and posts already in my tracked posts. Read the top 10 and pick the 3 whose commenters look most like my ICP, with one line on why for each. Check my capacity. Then stop and ask me before you track or qualify anything. When I say yes, track those posts, qualify their engagers and list the new qualified leads with ICP match, company and the comment or reaction behind each one."

Why it is written this way:

A time window and a dedupe rule. "Last 7 days" and "skip posts already tracked" keep the agent from bringing the same posts back each morning.
A small pick. Three good posts give you a list you can review over coffee. Ten mediocre ones give you a list you won't open.
A reason per post. If the agent has to say why a post fits, you can see when its judgment drifts.
An explicit stop. The agent reads your workspace's capacity and shows it to you before asking. Qualification needs your agreement, so the prompt makes the agent ask instead of guessing.

If your client runs the task while you are away, the run ends at the shortlist and the capacity check. You read it in the morning, reply "yes, go", and the agent tracks and qualifies the posts in the same chat. Processing can take a little time, so ask the agent to check the result before it repeats a request.

Swap the topics weekly if you like. Good topics are pains your product solves ("outbound reply rates dropping"), roles in transition ("first sales hire"), and comparisons people argue about in your category. For more ways to choose them, see how to find LinkedIn posts on any topic.

Finding this week's posts your buyers engage with

The quality of the daily list depends on the posts the agent picks. Engagement on a post is a signal of interest in a topic, and the right post brings you people who care about the problem you solve this week.

A few rules make the pick better:

Prefer posts about the problem over posts about you. A founder asking how others fixed their LinkedIn outreach draws people with the same problem. A post praising your category in general draws everyone.
Read the comments, not only the count. Ask the agent to quote two or three comments per post. Thoughtful comments from people in your market beat hundreds of likes from people outside it.
Mix your own and other people's posts. Your own posts, your team's posts, competitors' posts and creators in your niche all work. traxy qualifies the engagers the same way (find leads from any LinkedIn post).
Turn steady sources into Tracking. If the same creator or competitor keeps showing up in the agent's picks, track the profile instead. traxy checks the profile's recent posts when you add it, then follows a managed window of its five most recent posts published after that, so you stop finding them by hand (Tracking).

A like or comment shows interest in a topic, not a decision to buy. That is why every lead keeps the post, comment, like or repost behind it, with the date, and a Warm, Related or Activity label for how strong the evidence is. The ICP match score measures fit, not the chance someone buys.

If you'd rather not depend on posts at all, and if your plan includes Social Intent Search, the agent can search traxy's social intent database instead: plain-English search over 71M+ scanned LinkedIn likes, posts and comments, with the source and fit shown on every result.

Reviewing the leads and adding them to a campaign

Once the engagers are qualified, review them in the chat or on the Leads page. Then save the ones you want as a view and give that view to a campaign. Campaigns use shared lead views as their audience.

Read the summary. Ask: "List today's new qualified leads with ICP match, label and the comment behind each one." Check the source on anyone you plan to contact.
Clean the list. Mark leads good, bad, done or snoozed, add tags, and request an email or phone lookup where you need one. Lookups run only when you ask.
Save a view. "Save the Warm and Related leads from today's posts as a shared view called {name}." traxy_save_lead_view does this from chat.
Draft the campaign. "Draft a campaign for {view name}. Research each lead for {fact you care about}, group the leads by the post they came from and prepare a message variant for each group." A campaign can research every lead for the facts you choose before any message is prepared, then builds personalized LinkedIn and email sequences from that research (create a campaign).
Test it. The campaign runs on sample leads so you can check the research and drafts. Nothing is sent.
Review the launch. Check who will join, who is skipped because another unfinished campaign holds them, who a completed campaign contacted before, the sender, the message steps and the delivery mode (review and launch).

You don't need a new campaign every day. A common setup is one campaign per signal: save today's leads into the same shared view and run the campaign on it. You can run many campaigns at once on different signals, for example one for engagers on competitor posts and one for people commenting on posts about a pain you solve. For the full outreach side, see LinkedIn and email outbound with traxy.

What waits for your yes: capacity and sending

Everything that adds new leads or contacts a person waits for you. Reading never does.

Action	MCP tool	Needs your yes?
Find and read posts	`traxy_find_posts`, `traxy_read_post`	No. It reads posts
List and read leads	`traxy_list_leads`	No. It reads your leads
Track a post	`traxy_track_posts`	Yes. Ask the agent to show the posts before it adds them
Qualify a post's engagers	`traxy_qualify_linkedin_post`	Yes. You agree new qualified people may use capacity
Start a discovery run	`traxy_start_agent_run`	Yes. Same capacity agreement
Launch a campaign	`traxy_launch_campaign`	Yes. In chat, after the test and launch preview
Send a message or reply	`traxy_send_outreach_message`, `traxy_reply_inbox_conversation`	Yes. The AI shows the message and waits

The tools reference says traxy_start_agent_run and traxy_qualify_linkedin_post need a usage confirmation after you agree that new qualified people may use capacity. It also says a draft or preview does not authorize execution. From an AI chat, the agent shows what it will do and waits for your yes.

On the sending side, every customer can run campaigns from the traxy MCP: draft, test on sample leads, launch, manage sequences and senders, and reply in the inbox. LinkedIn and email go out only from your own connected sender accounts. Once you approve the launch, the campaign runs on autopilot, or you set it to approve every message in a review queue. Replies, unsubscribes and hard bounces stop the sequence, and all replies land in one inbox you can work from chat or the app. See LinkedIn GTM from your AI chat for the full MCP workflow.

Can traxy do this without an outside agent?

Yes, in part. If you don't run your own agent, trax, the agent inside traxy, can find posts on a topic ("Find posts about hiring a first sales team") and run discovery for new people who fit your ICP, with an approval card showing the proposed values and capacity before anything runs. trax also has a daily brief you turn on under Settings > Agent, delivered in traxy or Slack at the time you pick, and the Slack integration can send lead alerts and a daily summary. An outside agent adds the parts that live in your own setup: your schedule, your wording and the other tools it already uses.

What are the limits?

No schedule inside the MCP. The MCP gives an agent tools. When it runs is up to your client.
No unattended capacity use. Qualification and discovery runs wait for your agreement, so a fully unattended run stops at the shortlist.
Candidates are not leads. The scope shown before you confirm counts people to evaluate, not a promised number of qualified leads.
Fit is not a decision to buy. A 50%+ ICP match and an engagement mean the person fits and showed interest in a topic. Read the source before you write.
Signals outside LinkedIn come from research. Hiring, funding and job changes are checked by research on each lead in a campaign, not by a live event feed.
The agent only sees what you authorize. It works in the workspaces and with the tools your connection allows.

FAQ

Can an AI agent find LinkedIn leads for me every day?

Yes, through the traxy MCP. The agent finds posts on your topic, tracks the best ones and qualifies their engagers after you confirm capacity use. The daily schedule runs in your AI client, if it supports scheduled tasks; otherwise you run the prompt yourself each morning.

Which AI clients work with the traxy MCP?

Claude, ChatGPT, Codex, Cursor, Grok and other MCP clients. You add https://api.traxy.ai/api/v1/mcp and sign in to traxy once. No API key is needed.

Does the agent use capacity without asking?

No. Qualifying a post and starting a discovery run need your agreement that new qualified people may use your workspace capacity. Reading posts, leads and capacity does not use it.

Will the agent message the leads on its own?

No. Launching a campaign, sending a message and replying in the inbox all wait for your yes in the chat. After you approve a campaign launch, it runs on its own from your connected sender accounts, or you approve every message in a review queue.

How does traxy decide who counts as a lead?

traxy checks each engager against your ICP and keeps only people with a 50%+ match. Each lead shows the post, comment, like or repost behind it and a Warm, Related or Activity label.

Can I use one campaign for the leads the agent finds every day?

Yes. Save the qualified leads into a shared view and use that view as the campaign audience. You can also run several campaigns at once, one per signal.

How is this different from the trax daily brief?

trax is the agent inside traxy, and its daily brief arrives in traxy or Slack. The MCP lets your own agent, in the client you already use, run a lead search on your schedule, with your prompt.

Back to all articles