play-store-mcp
# Play Store MCP Server
[](https://github.com/lusky3/play-store-mcp/actions/workflows/ci.yml)
[](https://codecov.io/github/lusky3/play-store-mcp)
[](https://sonarcloud.io/summary/new_code?id=lusky3_play-store-mcp)
[](https://sonarcloud.io/summary/new_code?id=lusky3_play-store-mcp)
[](https://sonarcloud.io/summary/new_code?id=lusky3_play-store-mcp)
[](https://sonarcloud.io/summary/new_code?id=lusky3_play-store-mcp)
[](https://badge.fury.io/py/play-store-mcp)
[](https://github.com/lusky3/play-store-mcp/pkgs/container/play-store-mcp)
[](https://www.python.org/downloads/)
[](https://sonarcloud.io/summary/new_code?id=lusky3_play-store-mcp)
[](https://opensource.org/licenses/MIT)
[](https://glama.ai/mcp/servers/lusky3/play-store-mcp)
An MCP (Model Context Protocol) server that connects to the Google Play Developer API. Deploy apps, manage releases, respond to reviews, and monitor app health โ all through your AI assistant.
๐ **[Full Documentation](https://lusky3.github.io/play-store-mcp)**
## โจ Features
- ๐ **App Deployment** โ Deploy APK/AAB files to any track (internal, alpha, beta, production)
- โก **Batch Operations** โ Deploy to multiple tracks simultaneously
- ๐ **Multi-Language Support** โ Deploy with release notes in multiple languages
- โ
**Input Validation** โ Validate package names, tracks, and text before API calls
- ๐ **Automatic Retries** โ Built-in retry logic with exponential backoff for transient failures
- ๐ **Store Listings** โ Update app titles, descriptions, and videos for any language
- ๐ **Release Management** โ Promote releases between tracks, manage staged rollouts
- ๐ฅ **Tester Management** โ Add and manage testers for testing tracks
- โญ **Review Management** โ Fetch and reply to user reviews
- ๐ณ **Subscription Management** โ List subscriptions and check purchase status
- ๐ **In-App Products** โ List and manage in-app products
- ๐ฆ **Expansion Files** โ Manage APK expansion files for large apps
- ๐งพ **Orders** โ Retrieve detailed transaction information
- ๐ณ **Docker Support** โ Run as a container with health checks
- ๐ **Per-Request Credentials** โ Bring-your-own-credentials for multi-tenant deployments
- ๐ **Secure** โ Google Cloud service account authentication
## ๐ Quick Start
### Prerequisites
1. **Google Cloud Project** with the Google Play Developer API enabled
2. **Service Account** with access to your Play Console
3. **Python 3.11+**, `uvx`, or **Docker** installed
### Installation
#### Using uvx (Recommended)
```bash
# Run directly without installation
uvx play-store-mcp
```
#### Using pip
```bash
pip install play-store-mcp
play-store-mcp
```
#### Using Docker
```bash
docker run -e GOOGLE_APPLICATION_CREDENTIALS=/creds/key.json \
-v /path/to/service-account.json:/creds/key.json:ro \
ghcr.io/lusky3/play-store-mcp:latest
```
#### From source
```bash
git clone https://github.com/lusky3/play-store-mcp.git
cd play-store-mcp
pip install -e .
play-store-mcp
```
### Configuration
Set the path to your service account key:
```bash
export GOOGLE_APPLICATION_CREDENTIALS=/path/to/service-account.json
```
### Running with HTTP Transport
For remote access or public deployments, run the server with streamable-http transport:
```bash
play-store-mcp --transport streamable-http --host 0.0.0.0 --port 8000
```
The server exposes a `/health` endpoint for monitoring.
#### Per-Request Credentials (Recommended for Public Instances)
For public deployments where users bring their own credentials, configure your MCP client to pass credentials in headers:
```json
{
"mcpServers": {
"play-store": {
"url": "https://your-server.com/mcp",
"transport": "http",
"headers": {
"X-Google-Credentials-Base64": "YOUR_BASE64_ENCODED_CREDENTIALS"
}
}
}
}
```
To get your base64-encoded credentials:
```bash
base64 -w 0 < service-account.json
```
Per-request credentials are isolated โ each request uses only the credentials provided in its headers. No credentials are stored server-side or shared between requests.
#### Server-Side Credentials (For Private/Trusted Deployments)
For private deployments, set credentials via environment variable at server startup:
```bash
export GOOGLE_PLAY_STORE_CREDENTIALS='{"type":"service_account",...}'
# or
export GOOGLE_APPLICATION_CREDENTIALS=/path/to/service-account.json
play-store-mcp --transport streamable-http --host 0.0.0.0 --port 8000
```
### Read-Only Mode
To point the server at a live Play Console without any risk of mutating it, run
in read-only mode. All write tools (deploy, promote, halt, rollout, reply to
reviews, listing/tester updates) return an error instead of calling the API;
read tools are unaffected.
```bash
play-store-mcp --read-only
# or
export PLAY_STORE_MCP_READ_ONLY=1
```
### Code Mode (Experimental, enabled by default)
By default the tools are served as three meta-tools (`search`/`get_schema`/
`execute`) instead of the full tool list, cutting per-request tool-list token
overhead. The sandbox `execute` runs in is a base dependency, so this works
out of the box โ no extra install needed.
To opt out and use the classic tool list instead (`CODE_MODE` is env-only โ
there is no CLI flag):
```bash
export CODE_MODE=0
```
Under code mode one `execute` call can invoke up to 50 tool calls (including mutations) behind a single approval. Read-only enforcement still applies inside the sandbox, so pair it with `--read-only` / `PLAY_STORE_MCP_READ_ONLY=1` unless you need writes.
## ๐ง MCP Client Configuration
### Claude Desktop
Add to your `claude_desktop_config.json`:
```json
{
"mcpServers": {
"play-store": {
"command": "uvx",
"args": ["play-store-mcp"],
"env": {
"GOOGLE_APPLICATION_CREDENTIALS": "/path/to/service-account.json"
}
}
}
}
```
### Kiro
Add to `.kiro/settings/mcp.json`:
```json
{
"mcpServers": {
"play-store": {
"command": "uvx",
"args": ["play-store-mcp"],
"env": {
"GOOGLE_APPLICATION_CREDENTIALS": "/path/to/service-account.json"
}
}
}
}
```
### Gemini CLI / Other MCP Clients
```json
{
"mcpServers": {
"play-store": {
"command": "uvx",
"args": ["play-store-mcp"],
"env": {
"GOOGLE_APPLICATION_CREDENTIALS": "/path/to/service-account.json"
}
}
}
}
```
## ๐ ๏ธ Available Tools
### Publishing Tools
| Tool | Description |
| --- | --- |
| `deploy_app` | Deploy an APK/AAB to a track with optional staged rollout and single-language release notes |
| `deploy_app_multilang` | Deploy an APK/AAB with multi-language release notes |
| `promote_release` | Promote a release from one track to another |
| `get_releases` | Get release status for all tracks |
| `halt_release` | Halt a staged rollout |
| `update_rollout` | Update rollout percentage for a staged release |
| `get_app_details` | Get app metadata (title, description, etc.) |
### Store Listings Tools
| Tool | Description |
| --- | --- |
| `get_listing` | Get store listing for a specific language |
| `update_listing` | Update store listing (title, descriptions, video) |
| `list_all_listings` | List all store listings for all languages |
### Review Tools
| Tool | Description |
| --- | --- |
| `get_reviews` | Fetch recent reviews with optional filters |
| `reply_to_review` | Reply to a user review |
### Subscription Tools
| Tool | Description |
| --- | --- |
| `list_subscriptions` | List subscription products for an app |
| `get_subscription_status` | Check subscription purchase status |
| `list_voided_purchases` | List voided purchases |
### In-App Products Tools
| Tool | Description |
| --- | --- |
| `list_in_app_products` | List all in-app products for an app |
| `get_in_app_product` | Get details of a specific in-app product |
### Testers Management Tools
| Tool | Description |
| --- | --- |
| `get_testers` | Get testers for a specific testing track |
| `update_testers` | Update testers for a testing track |
### Orders Tools
| Tool | Description |
| --- | --- |
| `get_order` | Get detailed order/transaction information |
### Expansion Files Tools
| Tool | Description |
| --- | --- |
| `get_expansion_file` | Get APK expansion file information |
### Validation Tools
| Tool | Description |
| --- | --- |
| `validate_package_name` | Validate package name format |
| `validate_track` | Validate track name |
| `validate_listing_text` | Validate store listing text lengths |
### Batch Operations Tools
| Tool | Description |
| --- | --- |
| `batch_deploy` | Deploy to multiple tracks simultaneously |
## ๐ Google Cloud Setup
### 1. Create a Service Account
1. Go to [Google Cloud Console](https://console.cloud.google.com/)
2. Create a new project or select an existing one
3. Enable the **Google Play Developer API**
4. Go to **IAM & Admin** > **Service Accounts**
5. Create a new service account
6. Download the JSON key file
### 2. Grant Play Console Access
1. Go to [Google Play Console](https://play.google.com/console/)
2. Navigate to **Users and permissions**
3. Click **Invite new users**
4. Enter the service account email (from the JSON file)
5. Grant the following permissions:
- **Release apps to testing tracks** (for internal/alpha/beta)
- **Release apps to production** (for production releases)
- **Reply to reviews** (for review management)
- **View app information and download bulk reports** (for app details and orders)
## ๐ Environment Variables
| Variable | Description | Required |
| --- | --- | --- |
| `GOOGLE_APPLICATION_CREDENTIALS` | Path to service account JSON key | Yes (or use per-request credentials) |
| `GOOGLE_PLAY_STORE_CREDENTIALS` | Inline JSON credentials string | Alternative to file path |
| `PLAY_STORE_MCP_LOG_LEVEL` | Log level (DEBUG, INFO, WARNING, ERROR) | No (default: INFO) |
| `PLAY_STORE_MCP_DISABLE_DNS_REBINDING` | Disable DNS rebinding protection (for cloud/reverse-proxy deployments) | No |
| `PLAY_STORE_MCP_ADMIN_TOKEN` | Require `Authorization: Bearer <token>` on the `/credentials` endpoint (for deployments behind a reverse proxy) | No |
| `PLAY_STORE_MCP_READ_ONLY` | Disable all write operations (deploy, promote, rollout, reply, listing/tester updates) | No (default: off) |
| `PLAY_STORE_MCP_DOWNLOAD_DIR` | Directory that APK/AAB downloads are confined to (path-traversal / arbitrary-write protection). Downloads are always confined; defaults to the working directory when unset | No (defaults to cwd); **recommended** for network/hosted deployments โ the server warns if unset |
| `CODE_MODE` | Set to `0` to opt out of the code-mode transform and use the classic tool list | No (default: on) |
## ๐งช Development
### Setup
```bash
git clone https://github.com/lusky3/play-store-mcp.git
cd play-store-mcp
uv sync --dev
```
### Running Tests
```bash
uv run pytest -v --cov=src/play_store_mcp
```
### Linting
```bash
ruff check src/ tests/
ruff format src/ tests/
```
### Type Checking
```bash
mypy src/
```
## ๐ Troubleshooting
### Error: "Service account key not found"
Ensure `GOOGLE_APPLICATION_CREDENTIALS` points to a valid JSON file:
```bash
ls -la $GOOGLE_APPLICATION_CREDENTIALS
```
### Error: "The caller does not have permission"
Verify the service account has been granted access in Play Console with the required permissions.
### Error: "Package name not found"
Ensure the app exists in Play Console and the service account has access to it.
## ๐ License
MIT License โ see [LICENSE](LICENSE) for details.
## ๐ Acknowledgments
- Inspired by [antoniolg/play-store-mcp](https://github.com/antoniolg/play-store-mcp) (Kotlin)
- Built with the [MCP Python SDK](https://github.com/modelcontextprotocol/python-sdk)
- Uses the [Google Play Developer API](https://developers.google.com/android-publisher)
## ๐ค AI Usage Disclaimer
Portions of this codebase were generated with the assistance of Large Language Models (LLMs). All AI-generated code has been reviewed and tested to ensure quality and correctness.
TDQS
Scored across 3 tools
Each tool occupies a clearly separate stage of the workflow: search discovers tools, get_schema inspects parameters, and execute runs the actual call. There is no overlap or ambiguity between them.
Tool names are all lowercase and verb-focused, with get_schema following verb_noun while search and execute are bare verbs. The pattern is mostly predictable but not perfectly uniform.
Three tools is minimal but exactly matches the intended meta-workflow of discover, inspect, and execute. Each tool clearly earns its place, though the set is on the smaller end.
The workflow covers the full lifecycle from discovery to execution, enabling an agent to find and call tools effectively. A dedicated list-all-tools capability is missing, but search compensates for that gap.