PlugWiki
observability-mcp logo

Google Cloud Observability MCP Server: Setup for Claude, Cursor and Gemini CLI

OfficialLocal

Lets Claude, Cursor or Gemini CLI search your Google Cloud logs, read metrics and alerts, and look at traces and error groups.

This is Google's own server for Cloud Logging, Cloud Monitoring, Cloud Trace and Error Reporting. It runs on your computer and uses your Google sign-in from the gcloud tool, so the assistant sees only what your account can see. Every tool reads data and none changes it. Google calls the project a preview that is not an officially supported Google product, so expect breaking changes.

Good for

  • Developers on Google Cloud who debug production issues from the editor
  • On-call engineers who want logs, metrics and traces in one chat
  • Teams that want a read-only view of Cloud Logging and Monitoring

Works in

12 should work

Tap an app to see how to set it up and any known problems.

Install observability MCP

Claude Code

  1. 1

    Copy this prompt.

    Add the observability MCP server to Claude Code. Run this command:
    
    claude mcp add observability -- npx -y @google-cloud/observability-mcp
    
    Then check that it loaded and tell me if I need to sign in or restart.
  2. 2

    Paste it into Claude Code. It shows the change and asks before running it.

Requires Node.js, gcloud CLI signed in with gcloud auth application-default login.

Before first use, run gcloud auth login and gcloud auth application-default login. Enable the Cloud Logging, Monitoring, Trace and Error Reporting APIs you need in your quota project. Needs Node.js 20 or newer.

Built from the package data we hold. Not yet tested by us.

Tools

  • Searches your Google Cloud logs

    Searches log entries in a project with a filter, so the assistant can find error messages or follow what an app did.

    Try: “Show me ERROR logs from project my-app in the last hour.”

    Network · list_log_entries

  • Lists the logs in a project

    Lists the names of the logs a project has.

    Try: “Which logs exist in project my-app?”

    Network · list_log_names

  • Lists your log buckets

    Lists the log buckets in a project.

    Try: “List the log buckets in my-app.”

    Network · list_buckets

  • Lists views on your log buckets

    Lists the log views in a project.

    Try: “Which log views does my-app have?”

    Network · list_views

  • Lists where your logs are sent

    Lists the log sinks in a project, which route logs to other places.

    Try: “Where are my-app logs exported to?”

    Network · list_sinks

  • Lists log scopes

    Lists the log scopes in a project.

    Try: “List the log scopes in my-app.”

    Network · list_log_scopes

  • Finds which metrics a project records

    Lists the metric descriptors for a project so the assistant knows which metrics it can query.

    Try: “Which CPU metrics are available in my-app?”

    Network · list_metric_descriptors

  • Reads metric values over time

    Returns time series data for a metric you name, such as CPU use over the last hour.

    Try: “What was the average CPU use of my GCE instances in the last hour?”

    Network · list_time_series

  • Lists your alert policies

    Lists the alert policies set up in a project.

    Try: “What alert policies does my-app have?”

    Network · list_alert_policies

  • Lists alerts that fired

    Lists alerts in a project, newest first by default, with options to sort and page.

    Try: “Which alerts opened this week in my-app?”

    Network · list_alerts

  • Searches request traces

    Searches Cloud Trace for traces in a project.

    Try: “List traces from the past 30 minutes.”

    Network · list_traces

  • Opens one request trace

    Gets a single trace by its ID so the assistant can see where time was spent.

    Try: “Get trace 1a2b3c and tell me which call was slowest.”

    Network · get_trace

From the official docs

Permissions

low risk · 3 things to watch

Low risk

The assistant can read logs, metrics, alerts, traces and error reports in any Google Cloud project your account can see, and it cannot change anything. Use an account with viewer access only, since logs can hold private data.

What it can do

Watch out for

  • Logs can contain private data

    Log entries and traces may hold user data or secrets. Point it at projects you are fine sharing with your AI app. github.com

  • Uses your own Google permissions

    It reads whatever your signed-in account can read in the quota project and others. Use a sign-in with only viewer roles if you can. github.com

  • Preview software

    Google says it is not an officially supported product and may break when other tools change. github.com

Security advisories

No published security advisories as of Nov 1, 2026.

Technical details
Turning on read-onlyAll 13 tools are list or get calls; the server has no tool that writes or deletes.
Talks toGoogle Cloud Logging, Monitoring, Trace and Error Reporting APIs
Sign-in and keysUses Application Default Credentials set with gcloud auth application-default login. No key is stored by the server.

Questions

Is the Google Cloud Observability MCP server official?

It is published by Google in the googleapis GitHub organization. The README calls it a preview that is not an officially supported Google product and not covered by the Google Cloud Terms of Service.

What do I need before I install it?

Node.js 20 or newer and the gcloud CLI. Run gcloud auth login, then gcloud auth application-default login, and make sure the APIs you want (such as Cloud Logging) are enabled in your quota project.

Can it change or delete anything?

No. All 13 tools list or get data. It cannot write, update or delete.

Which AI apps does it work in?

The README covers Gemini CLI, Gemini Code Assist, Claude Desktop, Cline, Cursor and VS Code. Use the install box above for each one.

Why do I get API errors for some tools?

Each tool needs its API enabled in the quota project. For example, list_log_entries needs the Cloud Logging API enabled there. You can change the quota project with gcloud auth application-default set-quota-project.

How is it different from the gcloud MCP server?

The gcloud server runs gcloud commands. This one calls the Cloud Observability APIs directly and only lists or gets logs, metrics, alerts, traces and error groups. Both come from the same Google repo.

Information

updated Nov 2, 2026

Publisher
Official
Package
@google-cloud/observability-mcp
Registry
npm
Transport
stdio
Version
0.2.3
License
Apache-2.0
Last commit
Oct 17, 2026
Open issues
33

GitHubnpm

Sources

Docs, issue trackers and discussions this page is built from. Issues, quotes and findings above link to their own source.

Show all 3
  1. github.com · github.com
  2. github.com · github.com
  3. github.com · github.com

Facts from the official MCP registry, GitHub, npm, PyPI and the Docker catalog. Updated Nov 2, 2026.