One connection gives your assistant
lead generation, email verification and cold email campaigns
get_accountsearch_b2b_leadsHere are the profiles I found in the Getlead database:
The tab tax
Database, verifier, sequencer, inbox. Every campaign starts by opening all of them.
Export a CSV, clean it, re-upload it, map the columns, fix the ones that bounced.
From an idea to a launched sequence. On MCP it is one message and an approval.
01 · Source
Your assistant turns a plain-language ICP into database filters, samples the matches, and saves the ones that fit into a real Getlead list.
02 · Verify
Bulk verification runs a real SMTP handshake in the background and the assistant waits for it. Invalid and disposable addresses are dropped before your domain touches them.
03 · Personalize
The assistant reads the enrichment it just pulled, writes the opener around it, and shows you every message before one is queued.
04 · Launch
Campaigns run on your mailboxes with rotation, human-like delays and follow-ups. The assistant shows the sender, the audience and the sequence before anything queues.
05 · Optimize
Campaign analytics come back as structured data, so the assistant can rank the sequence, explain what is underperforming and rewrite it in place.
Works with
Claude and ChatGPT sign in with OAuth, no key to paste. Cursor and Gemini CLI take a single config block. Anything else that speaks MCP takes the same URL.
Setup
Settings, MCP in Getlead. Read-only for research, read and write when it should send.
Every client takes the same URL: https://mcp.getle.ad/mcp. No install, no proxy.
Claude and ChatGPT show a Getlead consent screen. Developer tools use the key header.
Describe the campaign. The assistant plans the steps and reports what it spent.
Copy-paste configs for every client live in the setup docs.
Guardrails
Enforced by the server, not suggested to the model.
Read-only for research and scheduled runs, read and write when it should send. Enforced at the server.
10 campaign launches and 30 replies per hour, on top of the plan rate limit.
Tool, client, status, credits and duration, readable in your dashboard.
Inbox content comes back fenced and marked untrusted, so a prospect cannot instruct your assistant.
MCP spends the same B2B, scraping and email allowances as the app.
Kill a key or disconnect a client and the very next call fails.
“My three SDRs share one license instead of three Apollo seats. 31 meetings last month and the data is cleaner.”
“Skeptical about the flat price, honestly. Landed my first client in 11 days and never looked at the competition again.”
“The niche filters are the secret. Tight list of 400 ops leaders, personalized opener, 6.4% reply rate on the first run.”
Pricing
No MCP add-on and no per-call fee. Tool calls spend the same B2B, scraping and email allowances you already pay for.
FAQ
MCP (Model Context Protocol) is the open standard for connecting an AI assistant to an external system. A remote MCP server exposes tools over HTTPS; the assistant reads the tool list and calls them on your behalf. Getlead's MCP server exposes the whole product: lead search, verification, lists, cold email campaigns, the unified inbox and the CRM.
Anything that speaks MCP. Claude (web, desktop, Claude Code), ChatGPT in developer mode, Cursor, Gemini CLI, Windsurf, Zed, Cline and custom agents built on the OpenAI or Anthropic SDKs. Claude and ChatGPT connect through OAuth; the developer tools use an API key header.
Pro and Ultra. The Starter plan does not include API or MCP access. If you connect on a lower plan the tools stay visible so your assistant can tell you what an upgrade unlocks, but calls are refused with an upgrade link.
No. A search through MCP spends exactly the same B2B credit it spends in the app, and a campaign spends your email allowance. There is no separate MCP quota and no per-call fee.
It can, which is why the guardrails matter. Tools that send are marked as write actions so your client asks for approval, the tool descriptions tell the model to confirm audience and copy first, and hourly caps limit campaign launches and replies. Use a read-only key if you want research only.
Both eras. It implements the stateless 2026-07-28 revision (server/discover, per-request metadata, mirrored headers) and still answers the older initialize handshake used by 2025-03-26 through 2025-11-25 clients, so nothing breaks while clients catch up.
Nowhere new. The MCP server runs inside Getlead's own API and reads the same workspace you see in the dashboard. Your assistant receives only what a tool returns, and mailbox credentials are never part of any response.
Revoke the API key or disconnect the OAuth client in Settings, MCP. Both take effect immediately, and the audit log keeps the history of what was called before that.
Two minutes to connect, then ask for leads, verification and campaigns in plain words.
Get started