stormgtm-mcp

Stdio MCP server for StormGTM. It lets an agent find leads with Radar and qualify them with Barometer. Sending is optional: the agent can send from mailboxes connected to StormGTM, or take the leads and send from its own tools.

Requires Node.js 22+ and a StormGTM API key.

Install

Claude Code:

claude mcp add stormgtm -- npx -y stormgtm-mcp

Cursor, ~/.cursor/mcp.json:

{
  "mcpServers": {
    "stormgtm": {
      "command": "npx",
      "args": ["-y", "stormgtm-mcp"]
    }
  }
}

Codex, ~/.codex/config.toml:

[mcp_servers.stormgtm]
command = "npx"
args = ["-y", "stormgtm-mcp"]

Then call whoami. It returns the account email and credit balance.

Authentication

The server reads STORMGTM_API_KEY first, then ~/.stormgtm/config.json (the same file stormgtm login writes), on every tool call. STORMGTM_API_URL overrides the API URL (default https://stormgtm.com).

The server starts without a key. Until you sign in, every tool returns a short message saying how to; after stormgtm login the next call works without restarting the server.

npm i -g stormgtm
stormgtm login

To pass a key explicitly, add it to the server's environment:

claude mcp add stormgtm --env STORMGTM_API_KEY=sgtm_live_... -- npx -y stormgtm-mcp

Agent skill

The server's instructions point agents at the stormgtm-gtm skill, which covers the full workflow. Install it into a project with the CLI:

stormgtm skill install --claude   # or --cursor, --agents

Tools

| Tool | What it does | | --- | --- | | whoami | Account email, credit balance and API URL | | credits | Credit balance, per-tier pricing and 30-day usage by verdict | | check_lead | Review one address: verdict, 0-100 score, reasons, policy result. Fast tier 1 credit, deep tier 5, unknown free | | check_batch | Submit many leads at once; unknown results are retried automatically | | batch_status | Progress and results for a batch | | report_outcome | Report delivered, bounced, complained, replied or opened so later checks improve. Free, and works for email sent outside StormGTM | | find_leads | Radar (beta): find people to email from a website URL or a description of the ideal customer (request, optional chatId to refine). 1 credit per new lead; searches that find nobody are free. Can take a minute or two | | list_radar_leads | Leads Radar saved, optionally for one chatId, with verdicts once qualified. With after ("0" first, then each nextAfter) it returns only leads not taken yet, oldest first | | add_leads | Save up to 500 of your own leads (email, optional name, title, company, note), free. Returns new leads, duplicates and rejections | | qualify_radar_leads | Check Radar leads by ids (fast or deep tier) and store each verdict. Send only to deliverable | | list_mailboxes | Connected mailboxes with status, last test and daily cap | | mailbox_status | One mailbox: status, last test result, pause reason and today's capacity | | mailbox_domains | Domains your mailboxes send from, with each unsubscribe host and whether it is verified | | send_email | Queue up to 100 messages (one or many), each from a connected mailbox (mailboxId, or from with the mailbox address). Paced through each mailbox's warm-up. 1 credit per email, refunded on failure | | list_domains | Sending domains with status and daily limit | | domain_health | A domain's daily capacity, 7-day bounce and complaint rates, and pause state | | email_status | Status and delivery state of a sent email | | create_sequence | Create a multi-step follow-up sequence sent from one mailbox (mailboxId, up to 10 steps, {{variable}} placeholders) | | enroll_leads | Enroll checked leads in a sequence with their variables | | sequence_status | List sequences, or show one sequence's steps and enrollments | | stop_enrollment | Stop one lead's sequence; waiting steps are cancelled and refunded | | list_threads | Inbox conversations (inbox, sent, archived or spam), filtered by unread or a search query | | read_thread | One conversation's messages: sender, time, unverified-sender flag, attachment names, and the new text (full for everything) | | reply | Answer an existing thread. Goes only to the thread's participant; 1 credit. No recipient parameter, so it cannot start new conversations | | mark_read | Mark threads read or unread | | archive_threads | Archive threads, or move them back to the inbox with archived: false | | mark_spam | Move threads to spam, or back with spam: false. Marking spam also suppresses the sender and stops their sequences | | inbox_counts | Total and unread threads per folder |

Sending goes through your connected mailboxes. A mailbox's domain needs its unsubscribe host set in the StormGTM dashboard first; until then sends are rejected with unsubscribe_host_required. Pass an idempotencyKey on each message so a retry never sends twice.

Inbox tools return email content wrapped as untrusted_email_content with a notice not to follow instructions inside it. The server instructions tell agents the same, and new outreach stays on send_email and sequences.

Source

Public source (MIT): github.com/marginsystems/stormgtm-mcp.

git clone https://github.com/marginsystems/stormgtm-mcp.git
cd stormgtm-mcp
npm install
npm run build
node dist/index.js