SEO Analytics MCP
# SEO Analytics MCP
MCP server for Google Search Console (GSC) + Google Analytics 4 (GA4).
## How It Works
### What `.venv` Means
- `.venv` is this project's private Python environment.
- It contains its own Python interpreter and dependencies for this repo only.
- `run-local.sh` activates `.venv` so the MCP server runs with the correct package versions.
### What `stdio` Means
- `stdio` means the MCP client and server communicate through standard input/output streams.
- The client starts `run-local.sh`, sends tool requests in, and reads JSON responses out.
- No web server or HTTP endpoint is required for the default local setup.
1. Your MCP client (Claude/Codex) launches `run-local.sh` as a stdio server process.
2. `run-local.sh` prepares `.venv`, installs dependencies, loads `config/.env`, and validates `config/service-account.json`.
3. The Python server (`seo-analytics-mcp`) receives tool calls, fetches data from GSC/GA4 using the service account, and computes merged analysis outputs.
4. Results are returned to the MCP client as structured JSON.
## Deployment Model
- This project is deployed as a local process on your machine via stdio.
- It is not an HTTP API service by default and does not require cloud hosting.
- To "deploy" it for another environment, clone the repo there, provide `config/.env` and `config/service-account.json`, then register that machine's `run-local.sh` path in the MCP client.
## Prerequisites
- Python 3.11+
## Quick Start
1. Create local config files from templates:
```bash
cp config/.env.example config/.env
cp config/service-account.example.json config/service-account.json
```
2. Fill required values:
- `config/.env`
- `DEFAULT_GSC_SITE_URL` (required format: `sc-domain:example.com`)
- `DEFAULT_GA4_PROPERTY_ID`
- `config/service-account.json`
- Replace all `REPLACE_WITH_...` fields with real Google service-account values.
3. Run the server:
```bash
./run-local.sh
```
The script creates/uses `.venv`, installs dependencies, loads `config/.env`, validates credentials, and starts `seo-analytics-mcp` over stdio.
## MCP Client Setup
Use the local launcher script as the MCP server command:
```text
/absolute/path/to/analytics/run-local.sh
```
### Codex App
Add server:
```bash
codex mcp add seo-analytics -- /absolute/path/to/analytics/run-local.sh
```
Verify:
```bash
codex mcp list
```
Remove:
```bash
codex mcp remove seo-analytics
```
### Claude Code CLI
Add server:
```bash
claude mcp add -s project seo-analytics -- /absolute/path/to/analytics/run-local.sh
```
Verify:
```bash
claude mcp list
```
Remove:
```bash
claude mcp remove seo-analytics
```
### Claude Desktop App (macOS)
1. Quit Claude Desktop completely.
2. Edit `~/Library/Application Support/Claude/claude_desktop_config.json`.
3. Add this server entry (merge into existing `mcpServers` if present):
```json
{
"mcpServers": {
"seo-analytics": {
"type": "stdio",
"command": "/absolute/path/to/analytics/run-local.sh",
"args": []
}
}
}
```
4. Start Claude Desktop again.
5. In a new chat, run tool `capabilities` to confirm connection.
## Main Tools
- `capabilities`
- `gsc_list_sites`
- `gsc_top_pages`
- `gsc_top_queries`
- `ga4_landing_pages`
- `analytics_merge_page_metrics`
- `analytics_generate_action_items`
- `analytics_trend_report`
- `analytics_data_quality_report`
## Screenshots
These screenshots are example outputs produced by using Claude Code with this MCP server.
The MCP provides the underlying GSC/GA4 data and analytics tools; dashboard rendering is not generated by default by the server itself.
For privacy reasons, values shown in the screenshots are randomized; the actual workflow was executed against real data.



## Troubleshooting
- `service-account.json still contains placeholder values`
- Replace all `REPLACE_WITH_...` fields in `config/service-account.json`.
- `No such file or directory: .../credentials/service-account.json`
- You have a stale credential env override from an older setup.
- Re-add the MCP server so it runs the latest `run-local.sh` and uses `config/service-account.json`.
- `Operation not permitted` when Claude Desktop starts MCP
- macOS is blocking execution from protected folders (commonly `~/Documents`).
- Either grant Claude Desktop access in `System Settings -> Privacy & Security` (`Files and Folders` and/or `Full Disk Access`) or move the repo to a non-protected path like `~/dev/analytics`.
- After changing permissions/path, restart Claude Desktop and re-add the MCP entry if needed.
- `Missing property_id and DEFAULT_GA4_PROPERTY_ID is not set`
- Set `DEFAULT_GA4_PROPERTY_ID` in `config/.env`.
- `Missing site_url...`
- Set `DEFAULT_GSC_SITE_URL` in `config/.env`.
- GSC permission error on `https://...` property
- This server enforces GSC properties in `sc-domain:<domain>` format to avoid URL-prefix mismatch issues.
- Use a Domain property in Search Console and grant your service account access there.
## Security
- Never commit `config/service-account.json` or `config/.env`.
- Commit only template files:
- `config/service-account.example.json`
- `config/.env.example`
TDQS
Scored across 16 tools
Most tools have distinct purposes, such as merging data, generating reports, or querying specific platforms. However, some overlap exists between analytics tools like 'analytics_popularity_snapshot' and 'gsc_top_pages', which could cause confusion in selection for similar tasks.
The naming is mixed, with prefixes like 'analytics_', 'ga4_', and 'gsc_' used consistently within groups, but the overall pattern is not uniform. Tools like 'capabilities' break the verb_noun convention, and there is no single naming style across all tools.
With 16 tools, the count is slightly high but reasonable for an SEO analytics domain that covers data merging, reporting, and platform-specific queries. It provides comprehensive coverage without being overly bloated.
The tool set offers complete coverage for SEO analytics, including data integration from GSC and GA4, trend analysis, opportunity identification, and action item generation. There are no obvious gaps, supporting end-to-end workflows effectively.