stormgtm

CLI and typed client for StormGTM. Barometer reviews an email lead before you send: give it an address plus whatever you know about the person, and get a verdict (deliverable, risky, undeliverable, unknown), a 0-100 score and plain-language reasons. Send the leads that pass from your own connected mailboxes.

Requires Node.js 22+.

Install

npm install -g stormgtm
stormgtm login
# the short alias runs the same binary:
sgtm whoami

Or the installer, which checks Node 22+ and runs npm install -g stormgtm:

curl -fsSL https://stormgtm.com/install.sh | bash

Add -s -- --mcp to also install the MCP server for agents.

Sign in

stormgtm login          # opens your browser, confirm the code, done
stormgtm login --key    # paste an sgtm_live_... key (input is hidden)
stormgtm whoami         # email, credits, key prefix, API URL (alias: status)
stormgtm logout         # remove the stored key
stormgtm keys           # open the API keys page in the dashboard
stormgtm config         # where the key and API URL come from
stormgtm config set api-url http://localhost:8787   # or: config unset api-url

The key is stored in ~/.stormgtm/config.json (mode 0600) as { "apiKey": "...", "apiUrl": "..." }. Environment variables win over the file:

| Variable | Meaning | | --- | --- | | STORMGTM_API_KEY | API key (sgtm_live_...) | | STORMGTM_API_URL | API base URL (default https://stormgtm.com) |

Commands

stormgtm check [email protected] --name "Jane Doe" --company Acme
stormgtm check [email protected] --deep --json
stormgtm batch leads.csv --wait
stormgtm batch-status <batch-id>
stormgtm outcome [email protected] bounced --detail "550 no such user"   # free; works for email sent by any means
stormgtm send --mailbox <mailbox-id> --to [email protected] --subject "Hello" --text "Hi Jane"
stormgtm send --from [email protected] --to [email protected] --subject "Hello" --text "Hi Jane"
stormgtm mailboxes
stormgtm mailbox-test <mailbox-id>
stormgtm mailbox-domains
stormgtm domains
stormgtm domain-health <domain-id>
stormgtm emails
stormgtm sequences
stormgtm sequence <sequence-id>
stormgtm enroll <sequence-id> leads.csv
stormgtm inbox --unread
stormgtm inbox --folder sent --search pricing
stormgtm thread <thread-id> --full
stormgtm reply <thread-id> --text "Thanks, here is the pricing."
echo "Thanks!" | stormgtm reply <thread-id>
stormgtm read <thread-id> <thread-id>
stormgtm unread <thread-id>
stormgtm archive <thread-id>
stormgtm unarchive <thread-id>
stormgtm spam <thread-id>
stormgtm unspam <thread-id>
stormgtm counts
stormgtm me

Radar (beta) finds people to email from a website or a description of your ideal customer:

stormgtm radar acme.io                       # progress on stderr, one lead per line on stdout, then a summary
stormgtm radar "CTOs at seed-stage dev tools startups" --chat <chat-id> --json
stormgtm leads [--chat <chat-id>]            # leads Radar saved, with verdicts once qualified
stormgtm leads --csv > leads.csv             # the same leads as a CSV file, free
stormgtm leads --after 0 --json              # only leads after a cursor, oldest first, as { leads, nextAfter }; pass nextAfter next time
stormgtm qualify-leads <lead-id> <lead-id> [--deep]
stormgtm add-leads leads.csv                 # your own leads, free; CSV with email plus name, title, company, notes

radar costs 1 credit per new lead. Leads you add with add-leads and searches that find nobody are free. A search needs at least 1 credit to start and stops when credits run out. radar exits 2 when it finds nobody. A search can take a minute or two. Qualify the leads before you send to them.

leads --csv prints name, email, title, company, company site, source URL, note, verdict and found date. Exporting is free. Cells that start with =, +, - or @ get a leading ' so a spreadsheet does not run them as formulas.

send goes out from a connected mailbox: pass --mailbox <id>, or --from with the mailbox address. Set each domain's unsubscribe host in the dashboard first; mailbox-domains shows where each one stands.

reply answers an existing thread only, goes to the thread's participant and costs 1 credit. In a terminal it asks before sending; pass --yes to skip the question.

check exits 2 when the lead is undeliverable, so it drops into shell pipelines.

| Exit code | Meaning | | --- | --- | | 0 | Success | | 1 | Error (API or network) | | 2 | Lead undeliverable, nothing was sent or enrolled, or Radar found no leads | | 3 | Usage error | | 4 | Not enough credits | | 6 | Not logged in, or the API key was rejected | | 7 | Rate limited | Add --json to any command for machine-readable output. CSV files for batch need an email column; name, company, companyDomain, title, githubLogin and other context columns are passed along.

Agent skills

stormgtm skill install --claude
stormgtm skill install --cursor --agents --json

Writes the stormgtm-gtm and stormgtm-send skills into .claude/skills/, .cursor/skills/ and/or .agents/skills/ of the current project (the git toplevel, else the current directory). Existing copies are overwritten; the command refuses to write through symlinks.

Library

import { StormGTM, clientFromEnv } from "stormgtm";

const stormgtm = clientFromEnv();
const result = await stormgtm.check({ email: "[email protected]", context: { company: "Acme" } });
if (result.verdict === "deliverable" && result.policy.allowed) {
  await stormgtm.send({ from: "Ada <[email protected]>", to: "[email protected]", subject: "Hello", text: "Hi Jane" });
}
await stormgtm.reportOutcome({ email: "[email protected]", kind: "delivered" });

clientFromEnv() reads the environment, then ~/.stormgtm/config.json.

Inbox:

const { threads, nextCursor } = await stormgtm.threads({ folder: "inbox", unread: true, q: "pricing" });
const thread = await stormgtm.thread(threads[0].id);
await stormgtm.reply(thread.id, { text: "Thanks, here is the pricing.", idempotencyKey: `reply-${thread.id}` });
await stormgtm.markRead([thread.id]);
await stormgtm.archiveThreads([thread.id]);
await stormgtm.spamThreads([thread.id]);
const counts = await stormgtm.inboxCounts();

Radar:

const { chatId, answer, leads } = await stormgtm.findLeads({
  content: "acme.io",
  onEvent: (event) => {
    if (event.type === "tool" && event.phase === "start") console.error(event.summary);
  },
});
const { leads: checked } = await stormgtm.qualifyRadarLeads(leads.map((lead) => lead.id));
const deliverable = checked.filter((lead) => lead.verdict === "deliverable");

findLeads creates a chat unless you pass chatId, streams the search and resolves when it finishes. It throws StormGTMError on an API error (for example insufficient_credits) or when the search fails. radarChats, createRadarChat, radarMessages, radarLeads, radarLeadsAfter (only leads after a cursor), deleteRadarLead and cancelRadarChat cover the rest of the Radar API.

MCP server

See stormgtm-mcp: claude mcp add stormgtm -- npx -y stormgtm-mcp.

Source

Public source (MIT): github.com/marginsystems/stormgtm-cli. Tags match npm versions.

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