Introduction
The GTM stack your agent is missing
Your AI agent can write outreach, schedule meetings, and qualify leads. But the moment it needs real data — a prospect’s LinkedIn URL, a verified email, evidence that a company is hiring SDRs, follower spikes on TikTok — it stalls. Stitching together RapidAPI calls, scraping rules, SMTP verifiers, and rate-limit logic is a project in itself.
GTM Tools is a monorepo of MCP servers that exposes that stack as a handful of tool calls. Every tool is metered in tokens, billed centrally, and reachable via MCP, REST, or CLI — so you focus on the agent loop, not the plumbing.
What is GTM Tools?
GTM Tools is a pnpm monorepo with four servers, each behind its own subdomain:
Each server is independent (its own deploy, its own tools.json) but shares one auth boundary (@mcp-tools/shared) and one billing wallet. New accounts get 100 free tokens to start.
Three ways in
MCP
Point an MCP-compatible client (Claude Desktop, Cursor, Windsurf) at the server URL with your bearer token. Tools appear in your agent’s tool list automatically.
What your agent can do
1. Provision auth and billing
get_api_key provisions a key via email verification. get_token_balance checks the wallet, buy_tokens tops up via Stripe (5 minimum), set_auto_reload charges the saved card when the balance dips below a threshold, and list_invoices returns the history. New accounts get 100 free tokens.
2. Find and act on LinkedIn
The most common starting point is get_linkedin_company_url — feed it a domain and it returns the company’s LinkedIn page. From there, list_linkedin_company_employees searches the workforce with boolean title filters ("(CEO OR CTO OR Founder) NOT intern"), get_linkedin_profile_url resolves a name + company to a profile, and get_linkedin_profile / get_linkedin_company fetch the full structured records. Once you’ve found the right person, send_linkedin_invitation and send_linkedin_message close the loop. connect_linkedin and list_connected_linkedin_accounts manage the underlying browser sessions.
3. Verify professional emails
get_email takes a name + domain and returns a deliverable email. It generates every common pattern (first@, first.last@, flast@, etc.) and verifies each against the domain’s mail server via SMTP — no mail actually sent. Catch-all domains are flagged in the response so your agent knows when to trust the result.
4. Detect buying signals
detect_signal runs every signal in one call. The individual detectors live alongside it: signal_socials_spike (Instagram/TikTok follower jumps), signal_hiring_role / signal_hiring_support / signal_hiring_sales_rep / signal_hiring_sales_leadership / signal_hiring_sales_rep_repost (job-board scans, including reposted SDR roles as a churn signal), signal_trustpilot_negative_reviews / signal_trustpilot_negative_support_reviews / signal_trustpilot_positive_reviews, and signal_technologies_identified (stack detection on the website). set_signals_order / get_signals_order configure execution order.
Quick example
Same flow from the CLI:
Available tools
Admin
Socials
Data
Signals
Packages
Hosted services
Dev mode
In dev mode, socials-tools and signals-tools run without WORKOS_API_KEY or STRIPE_SECRET_KEY. Auth is bypassed (every request is treated as org "dev") and token charges are skipped. Only admin-tools requires those keys, since it owns the WorkOS + Stripe integrations.