Every GTM tool your agent can use, and which one does the job.
293 tools counted · 144 official MCP servers · generated 2026-08-25 by build_directory.py (phase 1) · reconciled against tools_recount.py
Entry facts were pulled by hand: 261 on 2026-08-24, 32 on 2026-08-25. The generated date above is only the date this site was baked. Both dates ship because both rot. 277 of the 293 entries are unique products; 16 are the same product listed in a second category and counted once here.
An agent does not want a data enrichment tool. It wants a person's title from a LinkedIn URL. Type the thing you are trying to do. This runs in your browser over a baked index: no backend, no query logging, and it keeps working with the network cable pulled out.
Ordering is fixed and published, never tuned and never purchasable. official MCP first, then community, then unknown, then n/a, then none-found; within each band gate order is free, paid, enterprise-leaning, enterprise-only, unknown; then alphabetical by name. Computed, never curated, never purchasable. With a query typed, matches are banded by relevance first and the rule above breaks every tie. An exact name match pins to the top and nothing else is boosted. The filters above run over the 277 unique products, so their totals sit below the 293 entry totals used by the MCP status and gate views. Both numbers are correct and the difference is the 16 cross listings.
RevOps Infra is 22 of 23 reachable by an agent. Enablement & Coaching is 2 of 14. Gold is an official server, green is a community one. Every ratio below is read straight out of the category blocks in directory.json.
The server loads the baked file once at import and answers from memory. It makes zero outbound network requests, so it cannot be slow, cannot rate limit you, cannot cost anything, and cannot leak your query to a vendor. Everything network shaped happens in the weekly build.
{
"mcpServers": {
"gtm-directory": {
"command": "uvx",
"args": [
"gtm-mcp-directory"
]
}
}
}The package is not on PyPI yet, so this block is the shape the install will take rather than a working one-liner today. The server source is real and public: it lives in the gtm-mcp-directory repo and runs from a checkout right now.
Which tools can do X, and can your agent actually reach them. Filters on category, MCP status, access gate and tier, and states the ordering rule it used.
Every field on one entry, sources in full, plus the cross reference when the same product is listed in a second category.
The stat block. 293 entries, 144 official, 21 community, 117 none found, and 0 bench tested, scoped to a category when you ask for one.
The 15 category files with their counts, gates and source markdown, reconciled against tools_recount.py at build time.
Once a week: new entries, dead endpoints, gate changes, and every tool that moved between MCP statuses. Assembled from the machine output, never written from thin air.
PLACEHOLDER. The list runs on Kit and the form action is deliberately empty until that endpoint is set. No address is collected today and nothing is sent anywhere.
Anyone can submit a tool. Every submission is verified against public sources before it is listed, and the verification is the product. No vendor can pay to be listed, to rank higher, to be featured, or to soften a note. BENCH-TESTED cannot be bought at any price.
The submission queue is a GitHub issue form on the public gtm-mcp-directory repo. It is open now, and every submission is verified before it lists.
The questions people actually ask about GTM tools, MCP servers and agents, answered from this data with the numbers generated at build time and the date stamped on every one. No tool versus tool verdicts, because 0 tools here have been bench tested.
Every entry is RESEARCHED: facts from public sources with URLs, no usage claims, nobody has run the tool. BENCH-TESTED means Andrew personally ran it on a stated date. There are 0 of those, the number is published rather than hidden, and it is the proof the tier means something. 83 entries carry an access gate of unknown, 263 have no documentation URL recorded, and the 32 thinly sourced entries are named on the methodology page rather than quietly padded.