Multi Chat MCP Server (Google Chat)

Implements a pluggable provider architecture where each chat platform (Google Chat, Slack, Teams) is a self-contained module loaded dynamically at runtime via the --provider CLI flag. Each provider implements its own authentication, API layer, and tool registry following a standardized directory structure (src/providers/<platform_name>/), enabling concurrent execution of multiple providers and reducing coupling between platform implementations. The system uses Python's importlib to load provider modules and instantiate their mcp_instance.py entry points without modifying core server code.

Exposes chat platform operations as MCP tools using the @mcp.tool() decorator pattern, where each tool receives structured JSON parameters from AI assistants and returns typed results. The system translates natural language commands from MCP clients (Cursor, Claude Desktop) into tool invocations with validated parameter schemas, enabling AI assistants to programmatically interact with chat platforms without custom parsing logic. Tools are registered at server startup and advertised to clients via MCP's tool discovery protocol.

Implements production-ready Google Chat integration using OAuth 2.0 service account credentials stored in provider-config.yaml, with automatic token refresh and API error handling. The authentication system manages credentials lifecycle (loading, validation, refresh) and provides an authenticated HTTP client for all Google Chat API calls (messages, spaces, search). The provider abstracts Google's REST API behind a tool interface, handling pagination, rate limiting, and response transformation.

Implements a search_messages_tool that accepts natural language queries and optional date range filters, translating them into Google Chat API search requests with structured result ranking. The system supports date filtering patterns (e.g., 'last 7 days', 'since 2024-01-01') parsed from query parameters, and returns ranked results with message content, sender, timestamp, and space context. Search results are formatted for AI assistant consumption with relevance metadata.

Implements a send_message_tool that allows AI assistants to post messages to Google Chat spaces with optional thread context, preserving conversation structure and enabling threaded discussions. The tool accepts space ID/name, message text, and optional thread ID, translating them into Google Chat API calls that maintain message threading and space isolation. Results include message metadata (ID, timestamp, sender) for follow-up operations.

Implements tools for discovering Google Chat spaces accessible to the service account and listing space members with metadata (name, email, role). The system queries Google Chat API to enumerate spaces and members, returning structured results with display names, email addresses, and member roles. This enables AI assistants to understand team structure and identify relevant spaces for operations without manual configuration.

Implements a configuration system that loads provider settings from provider-config.yaml files and supports environment variable overrides for sensitive credentials (API keys, service account paths). The system validates configuration at startup, ensuring required fields are present and credentials are accessible, with clear error messages for missing or invalid configuration. Configuration is provider-specific, stored in src/providers/<platform>/provider-config.yaml, enabling per-provider customization without modifying core server code.

Provides integration guidance and configuration templates for connecting the MCP server to Cursor IDE and Claude Desktop via their respective MCP configuration files (~/.cursor/mcp.json, ~/.claude/mcp-server.json). The system documents how to register the MCP server as a tool provider in these clients, including command-line arguments for provider selection and credential passing. Integration is declarative via JSON configuration; no code changes required to clients.

Defines a standardized provider interface that new chat platform integrations must follow: directory structure (src/providers/<platform>/), entry point (mcp_instance.py), configuration file (provider-config.yaml), and tool registration pattern (@mcp.tool() decorators). The framework includes documentation and examples for adding new providers (Slack, Teams, Discord), with clear conventions for authentication, API abstraction, and tool definition. Developers can add providers without modifying core server code or understanding MCP protocol internals.

Provides debugging capabilities that trace natural language commands from AI assistants through the MCP protocol, provider selection, tool invocation, and API calls, with detailed logging at each stage. The system logs command intent, parameter parsing, tool selection, API requests/responses, and error context, enabling developers to understand how AI assistant commands map to chat platform operations. Debugging output is human-readable and includes diagnostic context for troubleshooting integration issues.