Introduction to Arclio MCP
Relevant source files
The following files were used as context for generating this wiki page:
Arclio MCP (Model Context Protocol) Tooling is a collection of standardized packages designed to enable AI models to interact securely and effectively with external services. Developed and maintained by Arclio, the project provides a bridge between Large Language Models (LLMs) and various enterprise systems such as Google Workspace, AWS S3, and Weaviate vector databases.
Sources: README.md:14-22, packages/google-workspace-mcp/README.md:38-40
The project leverages the Model Context Protocol to expose discrete, callable "tools" and "resources." This standardization ensures that AI assistants can discover and utilize external capabilities without direct exposure of sensitive API credentials to the model itself.
Sources: README.md:23-28, packages/aws-s3-mcp/README.md:38-42
Core Packages and Components
The monorepo is structured into specialized packages, each targeting a specific service ecosystem.
Arclio MCP Ecosystem Overview
The following diagram illustrates the relationship between the MCP Hub, the Arclio MCP Servers, and the external services they manage.
Key Packages
| Package | Description | Primary Capabilities |
|---|---|---|
google-workspace-mcp | Integration for Google services | Drive, Gmail, Calendar, Sheets, Slides |
aws-s3-mcp | Amazon S3 storage integration | Object listing, content retrieval |
weaviate-mcp | Vector database integration | Semantic search, hybrid search, CRUD |
markdowndeck | Technical foundation | Markdown-to-Google Slides conversion |
Sources: README.md:33-45, packages/markdowndeck/README.md:12-25
General Architecture
Arclio MCP servers follow a consistent execution flow to process requests from AI assistants. The server typically operates over standard input/output (stdio) and routes requests through a central application instance.
Request Execution Flow
- Discovery: The MCP Hub queries the server for available tools and resources based on configuration (e.g.,
GOOGLE_WORKSPACE_ENABLED_CAPABILITIES). - Routing: The
FastMCPapplication instance routes tool or resource calls to specific handler functions. - Service Interaction: Handlers call service-specific classes that interact with the external API (e.g., Google API, AWS SDK).
- Authentication: Service classes manage credentials (OAuth tokens, AWS Access Keys) internally.
- Response: Results are formatted and returned to the AI model.
Sources: packages/google-workspace-mcp/README.md:278-301, packages/aws-s3-mcp/README.md:46-51
Server Component Structure
The individual packages are structured to separate definitions from implementation logic:
Technical Foundation: MarkdownDeck
markdowndeck is an enterprise-grade library that serves as a foundational component for presentation generation within the Arclio ecosystem. It is utilized by the google-workspace-mcp package to handle complex slide creation from Markdown.
Pipeline Architecture
MarkdownDeck uses a four-stage pipeline to ensure predictable slide generation:
- Parse: Converts Markdown text into a structured
Deckobject. - Layout: A spatial planner that calculates positions and sizes for elements.
- Overflow: A recursive engine that partitions content across multiple slides if it exceeds boundaries.
- Generate: Produces the final JSON payloads for the Google Slides API.
Sources: packages/markdowndeck/README.md:65-80, packages/markdowndeck/README.md:120-125
Development and Configuration
The monorepo uses uv for workspace management and a Makefile for standardized tasks.
Environment Configuration
Servers are configured primarily through environment variables to manage security and feature flags.
| Variable | Description | Example |
|---|---|---|
GOOGLE_WORKSPACE_ENABLED_CAPABILITIES | JSON array of active services | ["drive", "gmail"] |
AWS_REGION | Target AWS region for S3 | us-east-1 |
WEAVIATE_URL | Host address for Weaviate | localhost |
WEAVIATE_OPENAI_API_KEY | Key for text vectorization | sk-... |
Sources: packages/google-workspace-mcp/README.md:118-124, packages/aws-s3-mcp/README.md:92-97, packages/weaviate-mcp/README.md:85-92
Common Development Tasks
The project provides a standardized set of commands for maintenance across all packages:
make setup-dev: Initializes the virtual environment.make install-all: Installs all workspace packages in development mode.make test: Executes the test suite (unit and integration).make lint/make format: Ensures code quality using Ruff.
Sources: README.md:52-108, pyproject.toml:60-65
Summary
Arclio MCP provides a robust, standardized framework for connecting AI assistants to various enterprise data sources. By utilizing the Model Context Protocol and the specialized markdowndeck pipeline, it ensures that AI interactions with external services are secure, scalable, and programmatically controlled.
Sources: README.md:14-22, packages/markdowndeck/README.md:20-25