tiger-cli

Tiger CLI exposes an MCP server that lets AI assistants interact with your Tiger Cloud infrastructure. The MCP server is started via the Tiger CLI using the embedded tiger mcp start workflow, and it exposes tools for managing and querying database services, as well as performing core database operations. Available tools include service_list, service_get, service_create, service_fork, service_start, service_stop, service_resize, service_update_password, service_logs, and db_execute_query, which enables parameterized SQL execution with support for timeouts and connection pooling. The server also proxies documentation-related tools from a remote MCP server, providing access to guides and documentation for PostgreSQL, TimescaleDB, and Tiger Cloud features. This proxy is enabled by default and can be disabled with a configuration flag.

To use the MCP tools, install Tiger CLI, authenticate, install/start the MCP server with tiger mcp install (or tiger mcp start after installation), and then interact with the tools via natural commands issued by your AI assistant or client. If your assistant supports the Streamable HTTP transport, you can configure the client with the appropriate endpoint as described in the manual installation section of the README. The MCP tools are designed to integrate with your existing Tiger Cloud authentication, so you will typically authenticate with tiger auth login first, then issue mcp-related commands through your assistant to manage services or run queries against your database services.

Prerequisites:

• A Tiger Cloud account and an authenticated CLI session
• Access to install Tiger CLI on your machine (macOS/Linux/WSL, Windows, or via package managers)

Installation options (choose one):

1. Install Script (macOS/Linux/WSL)

curl -fsSL https://cli.tigerdata.com | sh

2. Install Script (Windows)

irm https://cli.tigerdata.com/install.ps1 | iex

3. Homebrew (macOS/Linux)

brew install --cask timescale/tap/tiger-cli

4. Debian/Ubuntu (repository-based)

# Add repository
curl -s https://packagecloud.io/install/repositories/timescale/tiger-cli/script.deb.sh | sudo os=any dist=any bash

# Install tiger-cli
sudo apt-get install tiger-cli

5. Red Hat/Fedora (repository-based)

# Add repository
curl -s https://packagecloud.io/install/repositories/timescale/tiger-cli/script.rpm.sh | sudo os=rpm_any dist=rpm_any bash

# Install tiger-cli
sudo yum install tiger-cli

6. Go Install (from source)

go install github.com/timescale/tiger-cli/cmd/tiger@latest

After installing Tiger CLI, install the MCP server:

# Interactive installation (prompts for client selection)
tiger mcp install

# Or specify your MCP client directly, e.g. Claude Code
tiger mcp install claude-code

Manual installation (advanced): If your MCP client isn’t supported by tiger mcp install, follow the client's MCP setup instructions and start the server with:

# Example for a JSON-configured MCP server
# Save as a file (e.g., mcp_config.json) if your client requires it

To start using the Streamable HTTP transport instead of stdio, you can start the MCP server with:

tiger mcp start http --port 8080

And install it into your client using http://localhost:8080 as the URL.

Tips and common considerations:

• The MCP proxy feature provides access to documentation tools from the remote pg-aiguide server. This is enabled by default; disable it with: tiger config set docs_mcp false
• The MCP server uses your existing Tiger CLI authentication, so ensure tiger auth login has been completed prior to interacting with MCP tools
• The default configuration path for Tiger CLI is ~/.config/tiger/config.yaml, and you can inspect or modify settings with tiger config show/set/unset/reset
• If you run into transport issues, consider using the HTTP Streamable transport by starting the MCP server with the http transport and pointing your client to the provided URL
• The mcp_start command in the example uses the Tiger CLI as the host for MCP; this is suitable for local development and simple workflows, but you can adapt to other clients as needed
• When sharing MCP capabilities with AI assistants, ensure the assistant understands the available tools (service_list, service_get, service_create, service_start, etc.) and the database operation tool db_execute_query for executing SQL queries with proper parameterization and timeouts