Skip to main content
MCP server CX Codex

gcs — Claude Code MCP server: what it does, how to install

Google Cloud Storage integration for bucket and object tasks. Moves files directly between disk and cloud buckets without cluttering prompt context.

Install into Claude Code · Platform CLI All methods and platforms ↓
claude mcp add --env GCS_PROJECT_ID=<GCS_PROJECT_ID> --env GCS_SERVICE_ACCOUNT_KEY_FILE=<GCS_SERVICE_ACCOUNT_KEY_FILE> --env GCS_SERVICE_ACCOUNT_KEY_JSON=<GCS_SERVICE_ACCOUNT_KEY_JSON> --env GCS_BUCKET=<GCS_BUCKET> --env GCS_ENABLED_TOOLGROUPS=<GCS_ENABLED_TOOLGROUPS> --transport stdio gcs -- npx -y gcs-google-mcp-server
Open on GitHub

What gcs does

This server connects AI assistants directly to Google Cloud Storage. It handles inspecting buckets, reading text contents, transferring objects, and removing stale data. Tool availability is partitioned into readonly, readwrite, and delete permission sets, controlled via the GCS_ENABLED_TOOLGROUPS environment setting.

While simple inspection tools return inline strings, disk-oriented handlers stream raw bytes straight to local storage. Running download_prefix or put_object_from_path processes nested directory structures and binary files without forcing payloads through the LLM context window.

Setting GCS_BUCKET confines all file operations to a single bucket, while missing Google Cloud service account credentials will stop the server from launching.

What it helps with

  • stream binary files to Google Cloud Storage without token overhead
  • mirror remote storage prefixes directly to local folders
  • restrict agent capabilities to read-only bucket interactions
  • transfer objects between Google Cloud buckets

Example prompts for the agent

  • List all storage buckets accessible in the current Google Cloud project.
  • Upload the local archive dataset.zip into the bucket backups via streaming.
  • Download every object under the prefix exports/2024/ straight to local disk.

How to install gcs

Platform CLI

claude mcp add --env GCS_PROJECT_ID=<GCS_PROJECT_ID> --env GCS_SERVICE_ACCOUNT_KEY_FILE=<GCS_SERVICE_ACCOUNT_KEY_FILE> --env GCS_SERVICE_ACCOUNT_KEY_JSON=<GCS_SERVICE_ACCOUNT_KEY_JSON> --env GCS_BUCKET=<GCS_BUCKET> --env GCS_ENABLED_TOOLGROUPS=<GCS_ENABLED_TOOLGROUPS> --transport stdio gcs -- npx -y gcs-google-mcp-server

Replace <…> with your own values. 3 more environment variables are in the server README

Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more

Questions about gcs

How to install gcs in Claude Code?

Run `claude mcp add --env GCS_PROJECT_ID=<GCS_PROJECT_ID> --env GCS_SERVICE_ACCOUNT_KEY_FILE=<GCS_SERVICE_ACCOUNT_KEY_FILE> --env GCS_SERVICE_ACCOUNT_KEY_JSON=<GCS_SERVICE_ACCOUNT_KEY_JSON> --env GCS_BUCKET=<GCS_BUCKET> --env GCS_ENABLED_TOOLGROUPS=<GCS_ENABLED_TOOLGROUPS> --transport stdio gcs -- npx -y gcs-google-mcp-server`.

Why use disk streaming tools instead of inline reading?

Inline tools feed file contents through prompt context, which corrupts binary structures and exhausts token limits. Disk streaming writes raw bytes straight to the filesystem.

How do you block destructive actions?

Supply GCS_ENABLED_TOOLGROUPS="readonly,readwrite" in your environment config. This leaves delete operations disabled for the running session.

Similar MCP servers

All MCP servers for Cloud: AWS, Azure, GCP →

From the same set28

Project sponsors

Companies that keep this analytics open
Gold sponsor

Impressions, clicks and code copies — open statistics. We keep the numbers public for the same reason NeBlask itself is public. Become a sponsor