---
title: "Data vs. Intelligence"
description: "Why we don't just return raw API logs."
---
Most API integrations focus on **data delivery**: handing the computer a pile of numbers. This MCP focuses on **intelligence delivery**: handing the computer a set of findings.
## The Raw Data Problem
If you ask a model for keywords with "Striking Distance" (ranking 8-15) using raw data:
1. The agent calls the API.
2. It gets 5,000 keyword rows.
3. It tries to scan them.
4. It gets distracted or hits a context window limit.
5. It misses the keywords ranking at position 16.
## The Intelligence Solution
If you use the `seo_striking_distance` tool:
1. The MCP server fetches the 5,000 rows.
2. It applies a strict filter: `position >= 8 && position <= 15`.
3. It sorts them by `impressions` to find the most valuable ones.
4. It returns only the top 50 relevant keywords.
5. The agent receives a curated list of high-value opportunities.
## Deterministic Primitives
We categorize our tools into two types:
### 1. Data Proxies
Standard tools for listing sites, sitemaps, or running basic queries. High flexibility, low intelligence.
### 2. Intelligence Primitives (The "Pro" Tools)
Advanced tools that implement specific SEO logic:
* **Anomaly Detection:** Uses Z-scores to find statistically significant spikes or drops.
* **Trend Identification:** Compares two periods to find items with the most momentum.
* **Attribution Analysis:** Breaking down *where* a drop came from (mobile vs. desktop, specific countries).
By using intelligence primitives, you allow the AI agent to act like a senior SEO strategist rather than a data entry clerk.
