Why this matters
LLMs can generate insights from analytics data but lack a safe, standardized bridge to query production analytics. This repo implements a local MCP server that wraps Google Analytics Admin and Data APIs into explicit MCP tools, letting assistants such as Gemini or Claude run account/property lookups, reports, funnel and realtime queries without you building a bespoke connector.
What Sets It Apart
- MCP-native tooling: Provides explicit MCP tools (get_account_summaries, get_property_details, run_report, run_funnel_report, run_realtime_report, get_custom_dimensions_and_metrics) so an LLM can discover and call analytics operations as callable tools — which reduces prompt engineering and ad-hoc API glue.
- Designed for LLM integrations: Configuration examples target Gemini CLI/Code Assist and Claude Code, with a one-command pipx invocation pattern. So what: you can prototype conversational analytics workflows quickly with existing assistant tooling.
- Local control and credential model: Runs against Application Default Credentials (user OAuth or service account impersonation) and uses the Analytics Admin/Data APIs. So what: data access stays under your Google Cloud account boundaries and scopes (read-only recommended) rather than a third-party hosted connector.
Who It's For and Tradeoffs
Great fit if you want to prototype or demo LLM-driven analytics assistants, keep analytics data under your control, or integrate Google Analytics queries into an MCP-enabled agent workflow. Look elsewhere if you need a managed multi-tenant connector, high-throughput production hosting, or a UI-first analytics product — this repo is experimental, needs Python 3.10+, pipx, Google Cloud setup, and appropriate API scopes. Also note the server provides API access patterns rather than analytics interpretation: quality of insights still depends on prompts, model choice, and how you validate results.
