GEO Basics · Product Update

Every Mettle MCP tool, explained: brand profile, GSC data, GEO health, and more

A complete reference for the 23 tools the Mettle MCP server launched with (it now has more than 50) — what each one returns, which prompts get the most out of it, and how to combine get_brand_profile, get_gsc_data, and get_geo_health into a full GEO fix workflow in Claude Code.

Jun 11, 202612 min readBy the Mettle team

Reader guide

Turn this article into action

Plain-English point

A complete reference for the 23 tools the Mettle MCP server launched with (it now has more than 50) — what each one returns, which prompts get the most out of it, and how to combine get_brand_profile, get_gsc_data, and get_geo_health into a full GEO fix workflow in Claude Code.

Best use

Use this article when you are working on geo basics and need a concrete GEO next step.

Mettle MCP toolsMettle MCP serverGEO fix workflow Claude Codeget_brand_profile MCPget_gsc_data MCP

Update, October 2026: the server now exposes more than 50 tools. This post covers the original 23, which are all still available.

TL;DR: Mettle's MCP server exposed 23 tools at launch — from brand health and competitor intel to GSC cross-reference, GEO health audit, and your full brand profile. This post walks through every tool, what it returns, and which prompts get the most out of each one in Claude Code or ChatGPT.

What the Mettle MCP server does today

When you connect Mettle as an MCP server, your AI assistant gains direct read and write access to your GEO data. No screenshots. No copy-pasting numbers from a dashboard. No rebuilding context.

The server lives at https://api.mettlegeo.com/mcp and speaks the Model Context Protocol's Streamable HTTP transport. It works across Claude Code, Claude web, ChatGPT, and OpenAI Codex.

As of June 2026, the server exposes 23 tools across six categories:

CategoryTools
Account and brandlist_brands, get_brand_profile
Performance and healthget_brand_health, get_mention_trend, get_provider_breakdown, get_prompt_performance, get_geo_health
Competitor and citationget_competitor_intel, citation_list, get_topic_clusters
GSC and GA4 integrationget_gsc_data, get_ga4_data
Prompt campaignslist_prompt_sets, get_prompt_set_details, create_prompt_set, create_prompt, run_prompt_set, get_run_history, get_schedule_status
Alerts and learningcheck_alert_thresholds, get_llm_responses, tutorial
Client reportingget_client_report_data

This post covers three areas that most users have not fully used yet: the brand profile, the GSC cross-reference, and the GEO health audit.

Start here every time: get_brand_profile

Before any analysis session in Claude Code or ChatGPT, run this:

Use Mettle to get the brand profile for [brand].

get_brand_profile returns the full configuration for a brand: domain, industry, city, country, region, services or products, name variations, and search language.

This matters because every tool downstream reasons from this profile. If you ask Claude Code to generate a tailored llms.txt, write a prompt set, or suggest schema markup, it needs to know whether the brand is a product or a service company, what city it operates in, and what language its customers search in. Without the profile, the assistant guesses.

A full workflow in Claude Code:

Use Mettle to get the brand profile for [brand].
Based on that profile, generate a tailored llms.txt file for their domain.
Include their services, industry, city, and top focus areas.
Write it to llms.txt in the current directory.

That is a complete artifact the brand owner can drop into their /public folder. No manual work.

Key takeaway

get_brand_profile is the tool Claude Code needs before generating anything — prompts, schema markup, llms.txt, or an outreach brief. Always pull it first.

The most actionable signal: get_gsc_data

get_gsc_data is the newest tool and the most direct path to a concrete fix list.

It pulls your top Google Search Console queries, cross-references them with the AI search data Mettle has collected, and tells you which GSC queries have real search demand but no brand citation in AI responses.

Use Mettle to get GSC data for [brand]. Show me the top queries where we have impressions but no AI citation.

The tool returns:

  • summary — total GSC queries analyzed, how many have AI presence, how many cite the brand, and how many are missed
  • top_missed_queries — sorted by impressions, with click and position data, whether any AI is present at all, and which competitors appear in those responses
  • all_rows — the full cross-reference table for deeper analysis

The top_missed_queries list is where the work is. These are pages Google is already sending traffic to. AI systems are already active in those query clusters. Your brand just is not in the answer.

AvoidAsk the assistant to suggest general content improvements based on industry best practices.

PreferUse Mettle to pull GSC data, find queries with high impressions and no AI citation, then ask which content fix would close the biggest gap first.

A practical prompt for Claude Code:

Use Mettle to get GSC data for [brand].
Take the top 10 missed queries — high impressions, no brand citation.
For each one, suggest:
1. Whether the gap is a content gap, a schema gap, or a citation gap
2. The single most impactful fix
3. Which page on the site should be updated

This turns a raw data table into a prioritized sprint backlog. Claude Code can then open the relevant files, apply the fixes, and push a PR — all in one session.

Tip:

GSC data requires a connected Google Search Console account for the brand. If the tool returns connected: false, open the brand settings in the Mettle dashboard and complete the GSC OAuth flow first.

The diagnostic layer: get_geo_health

get_geo_health returns the most recent GEO Health Audit for a brand. If no audit has been run yet, it triggers one automatically.

Use Mettle to get the GEO health for [brand]. Summarize the top issues and the three most important fixes.

The audit scores the brand across five categories:

CategoryWeightWhat it checks
Technical SEO30%robots.txt, HTTPS, sitemap, schema markup, llms.txt
Content quality25%FAQ structure, content depth, heading hierarchy
Authority signals20%High-authority citations, domain trust signals
Freshness15%Content recency, update frequency
User signals10%Sentiment from AI-analyzed brand mentions

Beyond the score, the tool returns:

  • issues — a prioritized list of problems, each with severity and a fix description
  • improvement_roadmap — a phased plan organized into quick wins, medium-term work, and structural fixes
  • technical_details — specific booleans: whether robots.txt allows AI crawlers, whether an llms.txt file exists, which schema types were detected, whether HTTPS is active

In Claude Code, the schema detail is directly actionable:

Use Mettle to get the GEO health for [brand].
Look at schema_types_detected in the technical details.
If Organization, FAQPage, or Product schema is missing, generate the correct JSON-LD blocks
populated with the brand's real data from get_brand_profile.

Claude Code can write the output directly to a file. The brand owner pastes it into their CMS or template. Done.

Combining the three tools

The real power is combining get_brand_profile, get_gsc_data, and get_geo_health in a single session. Here is a complete diagnostic prompt:

You are a GEO specialist. Work through this in order:

1. Use Mettle to get the brand profile for [brand].
2. Use Mettle to get GSC data for [brand]. Identify the top 5 queries with high impressions and no AI citation.
3. Use Mettle to get the GEO health for [brand]. Note the overall score and the top 3 issues.

Then give me:
- The most likely reason AI systems are not citing this brand
- The single fix with the highest expected impact
- A concrete next step I can take today (specific page, specific change, specific file)

This is the workflow that would have taken three separate dashboard visits, multiple exports, and a full analysis session. With MCP it fits in one prompt.

The full tool reference

Account and brand

list_brands — lists all brands in the account with name, domain, and hash ID. Always run this first in a new session to get exact brand names.

get_brand_profile — returns the full brand configuration. Use before generating any brand-specific output.

Performance and health

get_brand_health — mention rate, sentiment, share of voice, and total mentions for the last 7 and 30 days. The core health check.

get_mention_trend — week-over-week direction: improving, declining, or flat. Includes the delta between the last two weeks.

get_provider_breakdown — mention rate and sentiment split by provider (OpenAI and Gemini). Shows which AI knows the brand best — and which ignores it.

get_prompt_performance — per-prompt mention rate and sentiment for the last 30 days. Shows which specific questions surface the brand and which do not.

get_geo_health — full GEO health audit: score, category breakdown, issues, schema details, roadmap.

Competitor and citation

get_competitor_intel — top competitors with their mention rates versus the brand's. Shows which providers each competitor dominates.

citation_list — domains that cite competitors but not the brand, sorted by competitor citation frequency.

get_topic_clusters — topics and keywords AI systems associate with the brand when they mention it. Useful for understanding how the brand is being framed.

GSC and GA4 integration

get_gsc_data — GSC top queries cross-referenced with AI citation data. Surfaces missed queries sorted by impressions.

get_ga4_data — GA4 totals, acquisition channels with revenue and conversion rate, top landing pages and AI-assistant referral traffic. Requires GA4 to be connected for the brand.

Prompt campaigns

list_prompt_sets — all campaigns for a brand with prompt count and run frequency.

get_prompt_set_details — all prompts in a set: query text, category, repetitions.

create_prompt_set — creates a new campaign. Supports daily or weekly schedule.

create_prompt — adds a new question to a campaign.

run_prompt_set — triggers an immediate run. Returns a run ID.

get_run_history — recent runs with status, timestamps, and prompt execution counts.

get_schedule_status — shows which campaigns run daily vs weekly and when they last ran.

Alerts and learning

check_alert_thresholds — pass/fail check against minimum mention rate, sentiment, or share of voice. Useful for automated monitoring scripts.

get_llm_responses — verbatim text responses from ChatGPT or Gemini about the brand. Shows exactly how AI systems describe the brand in their own words.

tutorial — explains Mettle concepts. Useful for onboarding new team members or clients who are new to GEO.

Client reporting

get_client_report_data — builds a client-ready AI visibility report and returns a short-lived link to the HTML report plus the headline figures, so an agent can add its own commentary.

What Claude Code can do with this

Claude Code with Mettle connected can complete full GEO fix workflows without leaving the terminal:

  1. Pull brand profile and GEO health
  2. Identify missing schema types
  3. Generate correct JSON-LD blocks from real brand data
  4. Write them to the right template file
  5. Pull GSC gaps and map them to existing pages
  6. Draft a llms.txt tailored to the brand
  7. Open a PR with all changes

No dashboard. No copying numbers. No rebuilding context.

That is the workflow the MCP tools are designed for. The analysis tools tell you what is wrong. The brand profile and GSC data tell you where it matters. Claude Code does the fix.

Setting up the connection

If you have not connected Mettle yet:

Claude Code:

claude mcp add --transport http mettle-mcp https://api.mettlegeo.com/mcp

OpenAI Codex — add to ~/.codex/config.toml:

[[mcpServers]]
name = "mettle-mcp"
transport = "http"
url = "https://api.mettlegeo.com/mcp"

ChatGPT or Claude web: add https://api.mettlegeo.com/mcp as a custom MCP connector in settings.

Verify with:

Use Mettle to list the brands in my account.

Sources and references

Questions this article answers

How many tools does the Mettle MCP server expose?

As of June 2026, the Mettle MCP server exposed 23 tools covering brand health, competitor intel, citation gaps, GSC cross-reference, GEO health audit, prompt campaign management, alerts, and verbatim LLM response retrieval. As of October 2026 it exposes more than 50.

What does get_brand_profile return?

get_brand_profile returns the full brand configuration: domain, industry, city, country, region, services or products, name variations, search language, and secondary language. It should be called before generating any brand-specific output such as llms.txt or schema markup.

What does get_gsc_data show?

get_gsc_data pulls your top Google Search Console queries, cross-references them with AI citation data, and surfaces which queries have high impressions but no brand citation in AI responses — the highest-priority content and schema fix opportunities.

What is the get_geo_health tool?

get_geo_health returns the latest GEO Health Audit for a brand: an overall score, category scores across technical SEO, content quality, authority, freshness, and user signals, plus a prioritized issue list and improvement roadmap. If no audit exists it triggers one automatically.

Can Claude Code fix GEO issues automatically using these tools?

Yes. Claude Code with Mettle connected can pull the brand profile, identify missing schema types from the GEO health audit, generate correct JSON-LD from real brand data, write the output to files, and open a pull request — all in one session.

See what AI engines say about your site

Run a Mettle audit to check whether ChatGPT, Gemini and Claude can read your pages, and where competitors get cited instead.

Run an audit