MCP Server

E-GEO integrates with the Model Context Protocol (MCP) to ground its analysis in real data instead of model guesses. When running through Claude Code, the pipeline uses MCP servers as its validation layer.

Why MCP matters for GEO

A GEO tool that claims competitor rankings without checking real search results is guessing. E-GEO’s rule: do not claim competitor rankings without search ground truth. MCP servers provide that ground truth.

Required servers for fully validated results

Server	Purpose	Criticality
brave-search	Market validation — competitor analysis, SERP data	High
chrome-devtools	Technical validation — rendered DOM, performance	High
fetch	Simple text scraping (fallback)	Medium

Behavior without MCP

E-GEO still runs if these servers are missing. It falls back to available tools (e.g. fetch / raw HTML) and clearly labels outputs as “Low Confidence”. You get results either way — you just know how much to trust them.

Setup

1. Add Brave Search and Chrome DevTools MCP servers to your Claude Code MCP configuration.
2. The serp collector (used by loop mode) also talks to the Brave API directly and needs BRAVE_API_KEY in the environment.
3. Run the validation-doctor skill — it checks your MCP dependencies and prints exact setup commands for anything missing:

npx skills add https://github.com/mverab/eGEOagents --skill validation-doctor

Validation flow

During a /geo <url> run:

1. Analyzer fetches the page (MCP fetch or web-reader fallback).
2. Ranker pulls real SERP data via brave-search to position your content against actual competitors.
3. Indexer can verify the rendered DOM via chrome-devtools — what AI crawlers actually see after JavaScript runs, not just raw HTML.
4. Any step that had to run without its preferred server marks its section of the report Low Confidence.

See also

• How It Works — where validation sits in the pipeline
• GEO Loop — the deterministic collectors that use the same data sources
• FAQ — troubleshooting MCP warnings