tagmanager-mcp
An MCP server for the Google Tag Manager API v2 that enables AI assistants to query GTM accounts, containers, tags, triggers, variables, and unpublished changes.
README
tagmanager-mcp
An MCP (Model Context Protocol) server for the Google Tag Manager API v2.
Ask your AI assistant about your GTM setup — accounts, containers, tags,
triggers, variables, unpublished changes — from Claude Code, Claude Desktop,
or any MCP client. Python, stdio transport, built on the official mcp SDK.
Why this one?
- Authenticate once — no re-auth treadmill. Auth is plain Google Application Default Credentials (ADC) with your own OAuth client: the refresh token does not expire, so you log in once and forget it. No hosted OAuth session that lapses every few days and demands another round of browser clicking.
- No service account required. The server runs as you, using the GTM permissions your Google account already has. There is no service-account JSON key to create, grant container access to, rotate, or accidentally commit.
- Local and direct. Runs on your machine over stdio; your GTM data flows
straight between you and
tagmanager.googleapis.com. No third-party proxy in the middle. - Built for LLM context windows. GTM's raw tag JSON is enormous (a single
GA4 event tag is easily hundreds of lines).
list_*tools return slim skeletons;get_*tools fetch full detail only when asked. - Quota-aware by design. The GTM API allows only 25 requests per 100 seconds per project. The server retries rate limits (429/403) and server errors with exponential backoff, and self-throttles after the first hit. Errors come back as actionable messages, not raw stack traces.
Tools (v0.1 — read-only)
| Tool | Purpose |
|---|---|
list_accounts |
GTM accounts you can access (optionally Google Tag accounts) |
list_containers |
Containers in an account |
list_workspaces |
Workspaces in a container |
get_workspace_status |
Unpublished changes and merge conflicts |
list_tags / get_tag |
Tags — skeleton list / full configuration |
list_triggers / get_trigger |
Triggers — skeleton list / full configuration |
list_variables / get_variable |
Variables — skeleton list / full configuration |
Every tool declares readOnlyHint, and the server requests only the
tagmanager.readonly OAuth scope. Write operations and publishing are
planned as separate, opt-in scope tiers (see Roadmap).
Prerequisites
- Python >= 3.10
- The gcloud CLI
- A Google account with access to your GTM containers
- Any GCP project you can enable an API on (used only for quota attribution)
Setup
1. Install
git clone https://github.com/<you>/tagmanager-mcp && cd tagmanager-mcp
python3 -m venv .venv
.venv/bin/pip install -e .
2. Enable the Tag Manager API on your quota project:
gcloud services enable tagmanager.googleapis.com --project=YOUR_PROJECT
3. Create a Desktop OAuth client (one-time, ~2 minutes).
Google blocks gcloud's built-in OAuth client for Tag Manager scopes ("This app is blocked"), so you bring your own:
- GCP Console → Google Auth Platform → Clients → Create client → Application type Desktop app → create, then download the JSON.
- On the Audience page, publish the app to Production. An app left in Testing status issues refresh tokens that expire after 7 days — the exact re-auth treadmill this project exists to avoid.
4. Log in
gcloud auth application-default login \
--client-id-file=path/to/your-client.json \
--scopes=https://www.googleapis.com/auth/tagmanager.readonly
gcloud auth application-default set-quota-project YOUR_PROJECT
The browser will warn "Google hasn't verified this app" — it is your own app; choose Advanced → Continue.
Already using ADC for other Google tools (BigQuery, analytics-mcp, ...)? Logging in replaces the ADC file, so include those scopes too, e.g.
--scopes=https://www.googleapis.com/auth/tagmanager.readonly,https://www.googleapis.com/auth/analytics.readonly,https://www.googleapis.com/auth/cloud-platform
Verify (expect HTTP 200 and your accounts):
curl -sS -H "Authorization: Bearer $(gcloud auth application-default print-access-token)" \
https://tagmanager.googleapis.com/tagmanager/v2/accounts
Connect an MCP client
Claude Code
claude mcp add gtm -- /absolute/path/to/tagmanager-mcp/.venv/bin/tagmanager-mcp
Claude Desktop (claude_desktop_config.json)
{
"mcpServers": {
"gtm": {
"command": "/absolute/path/to/tagmanager-mcp/.venv/bin/tagmanager-mcp"
}
}
}
Example prompts
- "Which GTM accounts and containers do I have?"
- "How many tags are in the default workspace of container GTM-XXXXXXX, grouped by type?"
- "Which tags are paused?"
- "Show me the full config of the purchase tag and which triggers fire it."
- "Does the current workspace have unpublished changes? What changed?"
- "Find triggers that no tag references."
Quota
The GTM API is tightly limited: 10,000 requests/day and 0.25 QPS (25 requests per 100-second window) per GCP project — per-user quota overrides do not raise it. Ordinary audit conversations fit comfortably; avoid "every tag in every container" sweeps across many containers at once.
Troubleshooting
- "This app is blocked" during login — you used gcloud's default OAuth
client; pass your own with
--client-id-file(Setup step 3). - 403 mentioning insufficient scopes — your ADC predates this setup; re-run the login command in Setup step 4.
- Errors mention enabling the API / quota project — run Setup step 2 and
set-quota-project; the error message itself carries the exact commands.
Development
.venv/bin/pip install -e ".[dev]"
.venv/bin/nox -s tests # stdlib unittest, fully offline
.venv/bin/nox -s lint # black --check
.venv/bin/mcp dev tagmanager_mcp/server.py # MCP Inspector
Roadmap
- v0.1 (current): read-only audit —
tagmanager.readonly - v0.2: create/update/delete for tags, triggers, variables, gated behind
explicit
confirm=truearguments — addstagmanager.edit.containers - v0.3: version creation and publishing, kept architecturally separate
from workspace editing — adds
tagmanager.edit.containerversionsandtagmanager.publish
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.