MCP Cookbook
The reference lists every tool. This is how to get the most out of it: compose Trakkr with the other MCP servers you already run, so a finding turns into a shipped fix in one conversation.
Read, act, then check the evidence
Start with a read from Trakkr. Use the evidence to draft a change in your editor, repository or CMS. Review the draft before committing, publishing or sending it. A later Trakkr read lets you compare measurements after the change.
Every recipe below is the same three-move loop:
Trakkr finds the gap and the fix
Your other MCP makes the change
Compare new measurements
Three reads to start with
Run these in order. Use an ID returned by list_brands in the next two calls. The JSON below uses a synthetic sample brand; replace its ID with your own.
All three calls are read only. Empty results are valid. Sample outputs are tested fixtures, not customer data.
1. Find a brand
{}{
"brands": [
{
"id": "00000000-0000-4000-8000-000000000001",
"name": "Sample brand",
"active": true,
"role": "viewer"
}
]
}Expected shape: brands. Parameters and constraints · HTTP response schema
2. Read citations
{
"brand_id": "00000000-0000-4000-8000-000000000001",
"view": "list",
"days": 30,
"limit": 10
}{
"brand": {
"id": "00000000-0000-4000-8000-000000000001",
"name": "Sample brand"
},
"citations": [],
"pagination": {
"total": 0,
"limit": 10,
"offset": 0,
"has_more": false
}
}Expected shape: brand, citations, pagination. Parameters and constraints · HTTP response schema
3. Read open actions
{
"brand_id": "00000000-0000-4000-8000-000000000001",
"status": "open",
"limit": 10
}{
"actions": [],
"pagination": {
"total": 0,
"limit": 10,
"offset": 0,
"has_more": false
}
}Expected shape: actions, pagination. Parameters and constraints · HTTP response schema
Recipes
Each one pairs Trakkr with a second MCP you already run. Jump to your tool, or browse by how you work. Connect the partner service and check your permissions before trying a recipe.
Post the Monday brief to your teamTrakkr + Slack · Team chat
GoalA standing AI-search ritual: five tight bullets in the channel every Monday.
Run /weekly-review for my brand and draft five bullets for #marketing in Slack. Show me the draft and ask before posting.
WhyThe weekly review chains visibility, the sharpest competitor move, the top citation gap and the quickest win into one scannable brief. Shared, it keeps the whole team pointed the same way.
Pair with a scheduler (a cron MCP or your assistant's scheduled tasks) to make it fire on its own each Monday.
Keep one brand book every AI readsTrakkr + Notion or Google Docs · Docs
GoalA single source of truth that steers every assistant, agency and writer to describe you correctly.
Pull my Trakkr brand-book resource and create (or update) a Notion page called "How to describe us" with it: what we are, how AI frames us today, what to set straight, and the proof.
WhyAI describes you from whatever it can find. A tight, shared brief, pasted wherever work happens, is the cheapest way to move how every model talks about you.
Turn this week's wins into ticketsTrakkr + Linear or Jira · Project tracker
GoalMake the fix queue real work, with owners and the brief attached.
Get this week's quick-win actions from Trakkr. For each one, create a Linear issue with the action title, why it matters, and the steps to take, and label them "ai-search".
WhyVisibility moves when the fixes actually ship. The action queue is already ranked by impact and effort, so the quick wins are the obvious first sprint.
Benchmark your brand against three rivalsTrakkr + Google Sheets · Scorecard
GoalA like-for-like view of who wins, on which prompts and models, with the evidence beside every conclusion.
Find my three closest AI-search competitors in Trakkr. Compare our visibility, rankings and perception over the last 30 days, then build a Google Sheet with one row per competitor, the biggest gap, the model where they win, and the prompt-level evidence I should inspect next.
WhyA competitor average can hide the useful part. Breaking the gap down by model, prompt and perception turns a league table into a concrete research queue.
Build a board update with proof attachedTrakkr + Google Slides · Executive reporting
GoalA short executive update that separates movement, shipped work and measured results.
Pull my latest Trakkr report, changes and results. Draft a six-slide Google Slides update covering: the headline movement, model-level winners and losses, competitor movement, work shipped, measured outcomes, and the next three decisions. Put the source date and supporting Trakkr evidence on every slide.
WhyThe report says what moved. Results compare measurements before and after the work. They do not establish causation. Keeping those claims separate makes an executive update useful and defensible.
Draft the citation-winning post in your CMSTrakkr + Webflow, Sanity or WordPress · CMS
GoalGo from a gap to a real draft, in the system your writers already use.
Use /content-brief for my brand on the topic "best running shoes for flat feet". Then create a draft post in Webflow with the outline as headings and the FAQ block in place, ready for my writer to finish.
WhyThe brief is built from the formats AI actually cites and your own brand voice, so the draft starts aimed at a real gap rather than a blank page. Evidence: trakkr://data/what-gets-cited
Find pages strong in Google but invisible to AITrakkr + Google Search Console · Search data
GoalBuild the upgrade list only you can see: pages that already rank in Google, that AI search still ignores.
Pull my top organic pages and their positions from Search Console. Cross-reference with Trakkr's citation data and tell me which pages rank well in Google but never get cited in AI search. Rank them by the size of the gap.
WhyAI models agree on who to recommend less than half the time, and they cite differently from how Google ranks. A page can win Google and still be invisible to ChatGPT. Evidence: trakkr://data/model-divergence
Needs Google Search Console connected to the brand in Trakkr.
Tie AI referral traffic to revenueTrakkr + PostHog or GA · Analytics
GoalCompare AI referrals with observed conversions, with the limits of attribution stated.
Get my AI referral traffic and sessions from Trakkr for the last 30 days. Pull conversions for the same window from PostHog, compare them by landing page, and show the observed referral and conversion counts. Flag gaps in the attribution data.
WhyCitations are the leading indicator; referred revenue is the lagging one. Tying them together is how AI search earns its budget line.
Needs Trakkr traffic tracking connected first (Settings, then your site script).
Find pages AI crawls but never citesTrakkr + Google Sheets · Content triage
GoalSeparate discovery problems from content problems, then rank the pages worth fixing first.
Get the pages AI crawlers visited in the last 30 days and cross-reference them with my citations. Put every crawled-but-never-cited page in a Google Sheet, add its Google search performance and Trakkr page analysis, then rank the top 20 by organic demand and fixability.
WhyCrawl and citation records show where to investigate. Missing telemetry does not prove a page was never fetched, and a fetch does not guarantee a citation. Inspect the page and the evidence before choosing a fix.
Needs crawler telemetry and Google Search Console connected to the brand in Trakkr.
Close a citation gap with a pull requestTrakkr + GitHub · Version control
GoalShip the page AI should be citing you for, as a reviewed pull request, without leaving the chat.
Find my top 3 AI citation gaps in Trakkr. Take the biggest one, draft a citation-optimised brief and the JSON-LD schema for it, then open a GitHub pull request that adds the page and the schema to my repo.
WhyReviews and comparison pages get cited far above their share of crawls, while product and help pages lose. Publish the format the data rewards, not another product page. Evidence: trakkr://data/what-gets-cited
Fix a page right on diskTrakkr + Filesystem · Local files
GoalFor a site that lives in a repo (Astro, Next, Hugo, Jekyll), rewrite a weak page in place with the exact fix.
Read my /pricing page from disk. Pull Trakkr's deep analysis of that URL, then rewrite the file in place: add the missing FAQ schema, tighten the answer copy, and keep my existing layout.
WhyClean, correct structured data tracks with getting cited. Light, accurate markup beats a kitchen-sink dump. Review any suggested markup against the page before adding it. Evidence: trakkr://data/schema-advantage
Audit AI crawler access and open the safe fixTrakkr + GitHub · Crawler access
GoalTurn a blocked crawler finding into a reviewable robots.txt or edge-config change without applying it blindly.
Audit AI crawler access for my brand in Trakkr. Show me the blocked agents, affected pages and current rule evidence. Preview the smallest safe fix, inspect my repository for the source robots.txt or edge rule, then open a GitHub pull request with that change. Do not merge it. Explain any crawler I should keep blocked and why.
WhyCrawler policy is not one allow-all switch. The useful fix preserves intentional training blocks while removing accidental search or retrieval blocks, with the exact rule reviewed before it ships.
Build your own recipe
The loop generalises. Any MCP server that can write something is a pair for Trakkr. Diagnose with a Trakkr tool or attach a trakkr:// resource, hand the result to the tool that does the work, then re-read Trakkr when new measurements are available. Record the change date, compare the same prompt and model coverage, and report missing evidence. A later movement alone does not prove the change caused it.
Want the full surface? The MCP server reference lists every tool, resource and slash-command workflow. The product tour shows it running inside Claude, ChatGPT and Cursor.