Slack MCP Server Tools

Build MCP servers that enable AI agents to interact with Slack APIs.

Problem Statement

AI agents need to perform actions in Slack beyond just sending messages - creating canvases, managing channels, posting rich content. MCP servers provide a standardized way to expose these capabilities as tools.

When to Use

• Building agent tools that create/update Slack canvases
• Exposing Slack API operations as MCP tools
• Returning rich resources (Block Kit, Work Objects) from tool calls
• Keywords: mcp server, slack tools, slack canvas, agent tools, model context protocol

Quick Start

module SlackCanvasMCPServer
  class Server < BaseMCPServer
    server_name "slack_canvas"
    server_version "1.0.0"

    tool :create_canvas
    tool :update_canvas

    def create_canvas(title:, content:, channel_id: nil)
      client = Slack::Web::Client.new(token: config[:slack_token])

      response = client.canvases_create(
        title: title,
        document_content: JSON.generate({ type: "markdown", markdown: content })
      )

      build_success_result(
        text: "Created canvas '#{title}'",
        canvas_id: response["canvas_id"]
      )
    end
  end
end

Architecture

Agent → Task → LLM → Tool Call → MCP Server → Slack API
                                      ↓
                              Tool Result with Resources

Key Patterns

Tool registration: Use tool :method_name DSL

Config injection: Credentials via config[:slack_token], config[:team_id]

Multi-channel results: Return ui:// + slack:// resources for different clients

Testing Strategy

RSpec.describe SlackCanvasMCPServer::Server do
  let(:server) { described_class.new(config: { slack_token: "xoxb-test" }) }

  it "creates canvas with valid content" do
    stub_slack_api(:canvases_create).to_return(canvas_id: "C123")

    result = server.create_canvas(title: "Test", content: "# Hello")

    expect(result[:canvas_id]).to eq("C123")
  end
end

Common Pitfalls

1. Token scopes: Ensure bot token has required scopes (canvases:write, etc.)
2. Rate limits: Handle Slack API rate limiting gracefully
3. Content format: Canvas content must be valid markdown or document JSON

Reference Files

• base-server.md - BaseMCPServer DSL and constants
• canvas-api.md - Canvas create/update/sections operations
• templates.md - YAML template system for resources
• multi-channel.md - Returning resources for multiple channels