Build a LinkedIn Campaign From ChatGPT or Claude
Salescadia's new MCP tool lets you build a LinkedIn campaign from ChatGPT or Claude: describe who to target and what you offer, then review and launch.
You can now build a LinkedIn campaign from ChatGPT or Claude without opening the Salescadia dashboard first. Connect your AI assistant to Salescadia's CRM MCP server, describe who you want to reach and what you're offering in one plain-English sentence, and the same engine that powers our in-app campaign wizard turns that sentence into a full outbound campaign — targeting, positioning, and a proven message cadence.
The point is not novelty for its own sake. Most reps already live in an AI assistant all day. Making a campaign should be as fast as describing it, and it should reuse the same targeting logic you'd get by clicking through the wizard — not a watered-down version.
This post covers exactly what the new create_campaign tool does, how to use it, and the guardrails that keep it from ever sending a message you didn't approve.
create_campaign is a write-scope MCP tool. It builds and configures a campaign; it does not source or send anything on its own. Every message still passes through your workspace's normal approval settings before it reaches a prospect.
What just shipped
Salescadia already exposes a Model Context Protocol (MCP) server so AI assistants like Claude, ChatGPT, and Cursor can read and act on your CRM. The read tools let an assistant look up contacts, deals, and campaign performance. The write tools let it actually do work — enroll leads, review a pre-screen queue, and now stand up a brand-new campaign.
create_campaign takes a plain-English description of your target audience and your offer, and runs it through the exact describeToFilters pipeline the campaign wizard uses. Your one sentence becomes:
| Your prompt supplies | Salescadia generates |
|---|---|
| Who to target | Titles, seniority, industries, locations, company sizes, and keyword groups |
| What you're selling | Campaign positioning and angle |
| Nothing extra | The proven default LinkedIn cadence, ready to run |
So a prompt like "create a campaign targeting VPs of Sales at US fintech startups, pitching faster lead routing" comes back as a structured campaign with strict keyword matching, seniority codes, and a message sequence — the same output you'd assemble by hand in the wizard.
How to build a LinkedIn campaign from ChatGPT (or Claude)
- Connect the MCP server. In Salescadia, generate an API key with the write scope and add Salescadia's MCP server to your assistant. The full tool list is published at salescadia.com/docs/mcp-tools.
- Describe the campaign. Tell your assistant who to target and what you're pitching, in a sentence or two. No filter syntax, no seniority codes to memorize.
- Review the draft. The tool returns a deep link straight to the new campaign in your dashboard. Open it, check the targeting and messaging, adjust anything.
- Launch when you're ready. Flip it live in the dashboard, or pass
launch=truein the original request to go live immediately. Once launched, lead sourcing begins within about 15 minutes.
That's it. The assistant does the assembly; you keep the final call.
Draft by default: nothing sends without your say-so
The reason a tool like this is safe to hand to an AI assistant is that it is deliberately conservative:
- Draft by default. A new campaign is created as a draft. Nothing is sourced and nothing is sent until you launch it.
- Approval settings still apply. Even with
launch=true, sending follows your workspace's existing approval rules. If you review messages before they go out, you still do. - Scope-gated. The tool is hidden from read-only API keys and refused if called with one. Read-only OAuth connectors stay read-only by policy, so a connected assistant can't quietly create outreach.
Already have a campaign running and just want to add a few people to it? Use add_leads_to_campaign instead — it drops specific LinkedIn profile URLs into an existing campaign, with the same "never re-touch someone already in flight" safety.
Why this matters
The friction in outbound is rarely the sending — it's the setup. Defining an audience, writing positioning, and building a cadence is the part reps put off. Collapsing that into a sentence you type into an assistant you already have open removes the main reason a campaign sits unbuilt.
It also keeps the work honest. Because create_campaign runs the same targeting and positioning engine as the wizard, an AI-built campaign isn't a lesser version — it's the identical logic reached through a faster door. The dashboard remains the source of truth, and your approval settings remain the gate.
That discipline is the same reason Salescadia's automation earns its keep. In one B2B sales case study, our routing and no-show protection combined to lift revenue 55% — $150,793 across 2,420 meetings — without adding headcount. You can read the full breakdown in the MedLeague case study. The throughline is consistent: put the mechanics on autopilot, keep the human in control of the decisions that matter.
See what better routing is worth to your pipeline
Estimate the revenue you're leaving on the table with a two-minute ROI calculation.
Run the numbersBuilding a campaign should take a sentence, not an afternoon. Connect your assistant, describe the campaign, review the draft, and launch when it's right.
More revenue. Same pipeline.