Connect tracKeting to your AI assistant
Ask Claude, Cursor or your coding assistant about your boards in plain English — brand rankings, ads, follower growth and YouTube AI coverage — and get your real tracKeting numbers back, right in the chat. No exports, no spreadsheets.
What it does
Once connected, you can just ask your assistant things like:
- "Rank the AI Board on X by views over the last year."
- "How is Anthropic's follower growth on LinkedIn trending?"
- "Which YouTube channels cover ChatGPT the most?"
- "What ads is Alex Hormozi running on Facebook?"
- "Show me Anthropic's best posts on X over the last month."
It answers using live data from your tracKeting account. Behind the scenes it can:
- List your boards — see exactly which boards your plan includes.
- Board totals — accounts, posts and ads on a board over the last 12 months (or channels, videos and views for YouTube AI coverage).
- Find an account — search a board by name or handle.
- Leaderboards — the top accounts on a board for any platform (X, LinkedIn, YouTube, Instagram, TikTok, Facebook), ranked by views. LinkedIn is ranked by likes instead, because LinkedIn never publishes view counts.
- Individual posts — what one account actually published on one platform over a stretch you choose: the words, the link, the numbers that platform reports, and the picture or video if you want it.
- Ads for one advertiser — the ads we hold for one account on Facebook (which covers Instagram), Google or LinkedIn, with the wording and the creative. You can narrow to one format, or to ads containing a particular word.
- Follower growth — the week-by-week follower trend for one account.
- YouTube AI coverage leaderboard — which AI brands get covered by the most channels and videos.
- Coverage trend — how one YouTube channel's AI-video output changed over the year.
- Brand coverage detail — the full YouTube coverage picture for one AI brand.
It's read-only — your assistant can look at the data, but it can't change anything in your account. And it only sees the boards you pay for: a Founders-only subscriber can't reach the AI Board tools, and vice-versa. Ask about a board that isn't on your plan and your assistant will say so and point you at plans — it won't quietly answer from a different board instead.
Connect it — pick your app
The apps split into two groups:
- Chat assistants (claude.ai, ChatGPT, Claude Desktop) — you just sign in to tracKeting when a window pops up. Nothing to copy or paste.
- Coding tools & editors (Claude Code, Cursor, VS Code, Codex, Windsurf, Cline, Zed) — you paste a token: a private key that links your account. Create one on your account page (you see it once), then drop it into the setup for your app.
Chat assistants — just sign in
No token to copy. When you connect, a tracKeting sign-in window pops up — sign in and approve, and you're done. The connection is read-only and limited to the boards you subscribe to.
Claude (claude.ai, on the web)
1. Open Settings → Connectors, then Add custom connector.
2. Paste this address:
https://mcp.tracketing.com/mcp
3. Leave the optional "Client ID" and "Client Secret" boxes empty — you don't need them. Click Add.
4. Find tracKeting in the list, click Connect, sign in when the window pops up, and press Authorize. Done — it now appears as a tool in your chats.
ChatGPT
1. In Settings, turn on Developer mode (it lives under "Connectors" or "Apps" — use the settings search if you don't see it).
2. Open Settings → Connectors, choose Create, and paste the address:
https://mcp.tracketing.com/mcp
3. For authentication choose OAuth and leave any client-ID / client-secret boxes empty, then Create and sign in when prompted.
Needs a paid ChatGPT plan and works on the web (not the mobile app). Official setup docs ↗
Claude Desktop (the Mac / Windows app)
1. Open Settings → Connectors, then Add custom connector.
2. Enter the address (name it tracketing):
https://mcp.tracketing.com/mcp
3. Leave the Advanced (Client ID / Secret) boxes empty, click Add, then Connect → sign in → Authorize.
Same sign-in as claude.ai — no token needed. Official setup docs ↗
Coding tools & editors — paste your token
Each of these uses the same token from your account page. Generate one (you see it once — it starts trk_live_), then swap it in below wherever it says YOUR_TOKEN.
Claude Code (the terminal app)
Paste this into your terminal, with your token in place of YOUR_TOKEN:
claude mcp add --transport http tracketing https://mcp.tracketing.com/mcp --scope user --header "Authorization: Bearer YOUR_TOKEN"
Then open a new Claude Code session. Claude Code connects its MCP servers when a session starts — so run the command first, then start a fresh claude session and the tracKeting tools appear right away. If you added it while a session was already running, that session won't see it until you restart. (--scope user makes it available in all your projects from then on — leave it off and it only works in the current folder.)
Your account page shows this exact command with your token already filled in. Official setup docs ↗
Cursor
Open Settings → Tools & MCP → New MCP Server (or edit ~/.cursor/mcp.json) and add:
{
"mcpServers": {
"tracketing": {
"url": "https://mcp.tracketing.com/mcp",
"headers": { "Authorization": "Bearer YOUR_TOKEN" }
}
}
}
Save, and tracKeting shows up in Cursor's tools list. Official setup docs ↗
VS Code (with GitHub Copilot)
Press Ctrl/Cmd+Shift+P, run MCP: Open User Configuration, and add this (note the top-level word is servers):
{
"servers": {
"tracketing": {
"type": "http",
"url": "https://mcp.tracketing.com/mcp",
"headers": { "Authorization": "Bearer YOUR_TOKEN" }
}
}
}
MCP tools run in Copilot's chat / agent view. Official setup docs ↗
Codex (OpenAI's terminal app)
1. Put your token in an environment variable named TRK_MCP_TOKEN — on macOS/Linux export TRK_MCP_TOKEN="YOUR_TOKEN", on Windows setx TRK_MCP_TOKEN "YOUR_TOKEN" (then open a new terminal).
2. Add the server:
codex mcp add tracketing --url https://mcp.tracketing.com/mcp --bearer-token-env-var TRK_MCP_TOKEN
Needs a recent Codex — run codex --version, and update with npm i -g @openai/codex@latest if --url isn't recognised. Official setup docs ↗
Gemini CLI
Add the server with your token (run in your terminal):
gemini mcp add --transport http --header "Authorization: Bearer YOUR_TOKEN" tracketing https://mcp.tracketing.com/mcp
Or add it straight to your settings file (~/.gemini/settings.json, or .gemini/settings.json in your project):
{
"mcpServers": {
"tracketing": {
"httpUrl": "https://mcp.tracketing.com/mcp",
"headers": { "Authorization": "Bearer YOUR_TOKEN" }
}
}
}
Note the field is httpUrl (not url) — that's the HTTP-streaming transport tracKeting uses. Run /mcp inside Gemini CLI to confirm it connected. Official setup docs ↗
Windsurf (now called Devin Desktop)
Open the MCP panel (the MCPs icon in Cascade, or Settings → Cascade → MCP Servers), edit the config file, and add:
{
"mcpServers": {
"tracketing": {
"serverUrl": "https://mcp.tracketing.com/mcp",
"headers": { "Authorization": "Bearer YOUR_TOKEN" }
}
}
}
Note the key here is serverUrl. Official setup docs ↗
Cline (the VS Code extension)
Click the MCP Servers icon → Configure MCP Servers (this opens the settings file), and add:
{
"mcpServers": {
"tracketing": {
"type": "streamableHttp",
"url": "https://mcp.tracketing.com/mcp",
"headers": { "Authorization": "Bearer YOUR_TOKEN" }
}
}
}
Use the settings file, not the simple "Remote Servers" box — that box has no place for your token. Official setup docs ↗
Zed
Open your settings (Ctrl/Cmd+, or run zed: open settings) and add:
{
"context_servers": {
"tracketing": {
"url": "https://mcp.tracketing.com/mcp",
"headers": { "Authorization": "Bearer YOUR_TOKEN" }
}
}
}
Zed calls MCP servers "context servers." A green dot next to tracketing in Settings → AI → MCP Servers means it connected. Official setup docs ↗
Getting the numbers right
Your assistant already knows these tips, but they're worth knowing yourself:
- Follower counts are point-in-time snapshots. For "how fast is this growing," ask for the follower-growth series, not a single latest number.
- Leaderboards and search show the head only. There's no "next page" — if a name isn't in the top results, narrow the search instead of paging.
- Totals cover the last 12 months. Board totals, rankings and coverage are all measured over the trailing year, not all time.
- Name the platform. Rankings and ads are per platform (X, LinkedIn, YouTube, Instagram, TikTok, Facebook) — say which one, or you'll get the default.
- Ads carry no performance figures. We record which ads ran, when, and what they said — never what they cost, or how many people clicked or saw them. So there's no "best performing ad", and an assistant that ranks ads for you has invented the ranking.
- Most LinkedIn ads have no dates. LinkedIn only publishes run dates for campaigns aimed at the EU, so about two in three come back without them. Those ads are real and they ran — we just can't say when, and we'd rather tell you that than guess.
- We hold posts from 6 July 2026 onwards. Ask about an earlier month and you'll get an empty answer for almost every account — that's where our records start, not a quiet patch. So ask about recent weeks, or don't name a period at all and let the answer tell you what it covers. Ads are the exception: for many advertisers we hold ads going back several months, and it differs from one to the next.
- Requests for individual posts and ads can cover up to three months at a time, and can't start before January 2025. If a stretch comes back empty, your assistant will tell you whether we simply recorded nothing then. Totals, rankings and follower trends cover the full history we hold.
Limits & what the messages mean
To keep things fair and fast, each account can pull:
- 3,000 rows of data per day and 300 requests per day (both reset at midnight UTC).
- Up to 50 results per answer, and a short pause if you fire off more than 20 requests in a minute.
- Of that daily allowance, 1,500 rows across 100 lookups can be individual posts and ads — roughly 20 accounts across three platforms every day. Everything else (totals, rankings, follower trends, search) draws on the main allowance only.
If you hit a limit, your assistant will show one of these — they're normal, not errors:
| Message | What it means |
|---|---|
| Daily row limit reached (3000). Resets at 00:00 UTC. | You've pulled your data for today. It resets at midnight UTC. |
| Daily call limit reached (300). Resets at 00:00 UTC. | You've made your requests for today. Resets at midnight UTC. |
| Rate limit: max 20 calls/minute… | You went a little fast — wait a few seconds and try again. |
| Daily limit for detailed post/ad rows reached (1500 rows)… | You've used today's allowance for individual posts and ads. Totals, rankings and trends still work. |
| Detailed post and ad lookups cannot start before 1 January 2025… | You asked for a stretch that starts earlier than we accept. Ask again from January 2025 onwards. |
What's covered
Four boards — the AI Board (AI brands), the Founders Board (indie founders), YouTube AI coverage, and Geneva real estate — across six social platforms (X, LinkedIn, YouTube, Instagram, TikTok, Facebook), month by month over the last 12 months. You'll only ever see the boards your subscription includes.
