You built the ICP. Now add the sending layer to your coding agent with one command.

The short answer for a GTM engineer

Yes, you can run LinkedIn outbound from a coding agent. Add the GTM Bud MCP server to Claude Code or Cursor, and the agent can create a campaign from an audience description, clone it for each segment, and return campaign results in the terminal. Launch waits for your confirmation of a preview, and the agent shows every reply for your review before it sends.

This page is for a GTM engineer or a technical RevOps person who works in Claude Code or Cursor most of the day. If you work in a chat app, start from the GTM Bud connector page.

Three situations this is built for

You built an ICP in another tool and need a sending layer. Your definition is good. The gap is the step after it. Here the handoff is a sentence: you give the agent the audience description, the agent writes the criteria from it, and GTM Bud checks them and finds the leads.

You were asked to clone a working campaign for five segments. This is a loop. The agent runs the loop, and you review five previews.

You want outbound results as data in the terminal. You ask, the tool returns results per campaign, and the agent writes them where your other GTM data lives.

Five prompts in your vocabulary

• “List my GTM Bud campaigns and write the results for each one to outbound/results.md as a table.”
• “Create a campaign draft for finance leaders (CFO, VP Finance, Controller) at B2B SaaS companies with 100 to 500 employees in the US and Canada that bill customers on usage. Show me the criteria check before you do anything else.”
• “Duplicate the SaaS finance campaign for these five segments from segments.yaml. Update the targeting for each copy and start a prospect preview for each.”
• “For every draft campaign, call the launch readiness check and give me one list of what is missing, grouped by campaign.”
• “Set the fintech campaign to follow-ups only. Tell me what changes before you apply it.”

A session transcript

This is a realistic session in Claude Code. The tool names are the real ones. The data is invented for the example, and results are left out because they would be your numbers, not ours.

you    Which campaigns do I have?
agent  list_campaigns
       2 campaigns: "SaaS finance leaders" (active), "Agency owners" (ended)

you    Draft a campaign for finance leaders at B2B SaaS companies with
       100 to 500 employees in the US and Canada that bill on usage.
agent  create_campaign_draft
       Draft created. Criteria check: role, size and region are clear.
       "Bill on usage" cannot be read from every profile. Suggested
       wording: "companies with usage-based pricing".
agent  update_campaign_targeting
       Targeting saved with the suggested wording.

you    Show me who matches.
agent  start_prospect_preview
agent  get_prospect_preview
       Real matching prospects returned for review.
agent  mark_prospect_fit
       2 marked Good fit, 1 marked Not a fit, as you said.

you    Clone it for the fintech segment.
agent  duplicate_campaign
agent  update_campaign_targeting
       Copy created with the fintech audience.

you    Write the offers and the sequence.
agent  get_value_offer_brief
agent  save_value_offers
       GTM Bud checked and saved the value offers.
agent  get_sequence_brief
agent  save_campaign_sequence
       GTM Bud checked and saved the message sequence.

you    Can it launch?
agent  get_launch_readiness
       Missing: no sending account is chosen. Choose it on gtmbud.com.
you    Done. Launch it.
agent  launch_campaign
       Launch preview shown. Waiting for your confirmation.
you    Confirmed.
agent  launch_campaign
       Campaign launched.

you    Results for the SaaS campaign, then stop new first messages on it.
agent  get_campaign_results
agent  get_campaign_controls
agent  set_sending_mode  (mode: followups)
       Applied. Only follow-ups to messages already sent will go out.

A worked example: one segment in detail

The company in this example sells billing reconciliation software. The names are invented.

The audience sentence. “Finance leaders (CFO, VP Finance, Controller) at B2B SaaS companies with 100 to 500 employees in the US and Canada that bill customers on usage.”

What the criteria check said. Role, company size and region were clear. The qualifier “bill customers on usage” could not be checked for every company, so the check flagged it as hard to verify from public data and suggested “companies with usage-based pricing”.

Two previewed prospects.

• Elena V., VP Finance at a 240-person API infrastructure company in Toronto with public usage-based pricing. Marked Good fit.
• Tom H., Controller at a 150-person HR software company in Austin that sells flat annual seats. Marked Not a fit: right role and size, wrong pricing model.

One value offer. “A sample reconciliation of one month of your usage invoices against your ledger, with every mismatch listed.” The agent read the GTM Bud instructions, researched the company website, wrote the offer, and GTM Bud checked and saved it.

One reply draft. A lead wrote: “We do this in spreadsheets today. How long does the sample take?” The agent wrote, for review:

“The sample takes about a week once we have one month of invoices and the matching ledger export. You get a list of every mismatch, with or without a call after. Should I send the two file formats we need?”

The engineer approved the text, and the agent sent that exact text.

The results question. “Give me results for the SaaS campaign and the fintech copy side by side, and append them to outbound/results.md with today’s date.”

Three objections GTM engineers raise

“I already have the list. Why can I not push it in?” Because the connector does not take a list upload. Targeting is an audience description that your agent turns into criteria. GTM Bud checks the criteria and uses them to find leads. That costs you control of the exact rows. It gives you a criteria check and a preview of real matches before any message goes out. If your process depends on a fixed list from another system, this is the wrong layer for that step.

“I want this unattended, on a schedule.” The connector is built for a session with you in it. Launch needs your confirmation of a preview. Each change that starts sending needs your confirmation. The agent shows each reply for your review, and GTM Bud instructs it to send only after you approve the exact text, up to 40 replies a day for each LinkedIn account. A change that stops sending applies at once. Plan for an agent that prepares everything and a person who says yes.

“Half the setup is still in a web app.” True for three things: connecting LinkedIn accounts, choosing the sending accounts for a campaign, and setting the sending schedule are done on gtmbud.com. The readiness check tells the agent which of these is missing, so you find out in the terminal and fix it in the browser once.

Setup and the full capability list

Setup is one command, claude mcp add --transport http gtm-bud https://gtmbud.com/mcp, and the steps for Cursor and Codex are on GTM Bud for coding agents. For everything the connector can do, see the GTM Bud connector page.