slack-relay-mcp

Implements the Model Context Protocol (MCP) as a transport layer to relay Slack messages and events into Claude or other LLM clients. Uses MCP's resource and tool abstractions to expose Slack API operations (send, read, list messages) as standardized LLM-callable functions, enabling bidirectional Slack integration without direct API key exposure to the LLM.

Exposes a command-line interface for sending messages to Slack channels and retrieving message history without requiring LLM integration. Uses Node.js child process execution to invoke Slack API calls, supporting both synchronous message sends and asynchronous channel history queries with optional filtering by timestamp or user.

Exports a typed TypeScript/JavaScript API for Slack operations (send message, list channels, get message history, post reactions) with full type definitions and async/await support. Wraps the Slack Web API client with convenience methods that handle pagination, error handling, and response normalization, enabling type-safe Slack integration in Node.js applications.

Registers slack-relay-mcp as a Claude Code skill, allowing Claude's code interpreter to invoke Slack operations directly during code execution. When Claude writes or executes code, it can call Slack functions to send messages, read channels, or retrieve history as part of a multi-step reasoning workflow, with results fed back into Claude's context for further analysis.

Exposes Slack channels and messages as MCP resources (read-only or read-write), allowing LLM clients to browse and reference Slack data through the MCP resource protocol. Resources are identified by URIs (e.g., slack://channel/C123456) and return structured JSON representations of channels, message threads, and user metadata, enabling LLMs to understand Slack context without making direct API calls.

Defines MCP tool schemas for Slack operations (send_message, get_channel_history, list_channels, add_reaction) that LLM clients can invoke through the MCP function-calling protocol. Each tool includes input validation schemas, error handling, and response normalization, allowing LLMs to call Slack operations with type-safe arguments and receive structured results.

Handles Slack bot token authentication through environment variables or configuration files, managing credentials securely without exposing them to LLM contexts. Uses the @slack/web-api client under the hood to authenticate with Slack's OAuth 2.0 flow, supporting token rotation and scope validation to ensure the bot has required permissions.

Implements cursor-based pagination for Slack message history retrieval, allowing efficient querying of large channels without loading all messages into memory. Supports filtering by timestamp range, user ID, or message type, with automatic cursor management and result normalization to handle Slack API's pagination format.

Implements exponential backoff retry logic for transient Slack API errors (rate limits, temporary outages), with configurable retry counts and backoff strategies. Distinguishes between retryable errors (429 Too Many Requests, 5xx) and permanent errors (401 Unauthorized, 404 Not Found), providing detailed error messages and logging for debugging.

Supports connecting to multiple Slack workspaces simultaneously by maintaining separate bot token credentials and routing operations to the correct workspace based on channel or user identifiers. Enables a single MCP server instance to serve as a bridge for multiple Slack workspaces, with workspace-aware resource URIs and tool parameters.