Developer Docs

MCP Server

Connect AI assistants like Claude, Cursor, and other MCP-compatible tools directly to your Geolix data.

Overview

Connect AI assistants like Claude, Cursor directly to your Geolix data.

Capabilities

📈

Brand Visibility

Check brand visibility across ChatGPT, Perplexity, Gemini, and more.

⚔️

Competitor Compare

Compare competitors on visibility, sentiment, share of voice.

🔗

Source Analysis

See which domains and URLs AI models cite most often.

Content Inspection

Inspect what AI engines actually read.

Setup

1

Create an API Key

Log in to app.geolix.ai,go to Settings → API Keys,create a key and copy it.

2

Connect to Your AI Tool

PropertyValue
Server URLhttps://api.geolix.ai/mcp
TransportStreamable HTTP
Authenticationx-api-key header
3

Start Asking Questions

Try these natural language prompts:

  • "How has our brand visibility changed over the past 30 days?"
  • "Which competitors have the highest share of voice in ChatGPT?"
  • "Which URLs are cited most by AI search engines?"

Supported Platforms

Claude Code

claude mcp add geolix --transport http https://api.geolix.ai/mcp \
  --header "x-api-key: YOUR_API_KEY"

Cursor

SettingsTools & IntegrationsMCPAdd Custom MCP

VS Code (GitHub Copilot)

Create .vscode/mcp.json:

{
  "mcpServers": {
    "geolix": {
      "type": "http",
      "url": "https://api.geolix.ai/mcp",
      "headers": {
        "x-api-key": "${GEOLIX_API_KEY}"
      },
      "timeout": 120000
    }
  }
}

Windsurf

SettingsMCPAdd Server

How it works

AI Assistant  ─MCP protocol──▶  Geolix MCP Server (:8101)
                                   │
                                   ├── list_tools()  → returns 28 tool definitions
                                   │
                                   └── call_tool()   → direct service layer call
                                                        │
                                                        ▼
                                                   Database → JSON → MCP response

The MCP Server is a standalone FastMCP process with manually defined tools.

Permissions

Most tools are read-only — dashboard, sources, detection records only pull data.

A few tools modify configuration(create brands, edit intents)。Write tools are marked with readOnlyHint: false,delete tools with destructiveHint: true;AI 客户端执行前会请求确认。

Write operations require owner or admin role. Any project member can use read tools.

Tools

The server currently provides 28 MCP tools:

DomainCountDescription
brands4Brand management
projects8Project management
queries2Intent management
runs4Detection runs
board9Dashboard data
insights1Analytics insights

Brands

ToolR/WDescription
list_brandsReadList all brands
get_brandReadGet brand details
create_brandWriteCreate a brand
update_brandWriteUpdate a brand

Projects

ToolR/WDescription
list_projectsReadList all projects
get_projectReadGet project details
create_projectWriteCreate a project
update_projectWriteUpdate project settings
list_queriesReadList monitoring intents
add_queriesWriteAdd monitoring intents
suggest_topicsReadAI-generate monitoring topics
suggest_queriesReadAI-generate intent suggestions

Runs

ToolR/WDescription
get_runReadGet detection run details
list_runsReadList detection runs
trigger_runWriteTrigger a detection run
cancel_runWriteCancel a detection run

Board

ToolR/WDescription
get_boardReadDashboard overview
get_board_sourcesReadCitation sources
get_source_detailReadSource detail
get_competitive_gapReadCompetitive gap analysis
get_by_intentReadMetrics by intent
get_chat_recordsReadQ&A records
get_chat_facetsReadFilter facets
export_chatsReadExport Q&A records
get_completionReadCompletion status

Queries

ToolR/WDescription
update_queryWriteUpdate a monitoring intent
archive_queryWriteArchive a monitoring intent

Others

ToolR/WDescription
reanalyze_projectWriteRe-analyze project data

Troubleshooting

401 Unauthorized

403 Forbidden

Connection timeout