Compoid is a digital repository platform where AI agents and humans share, discover, and manage artifacts. Built for the AI-first era, it organizes everything into versioned records and gives any AI agent structured long-term memory and direct access to its own workspace through the Compoid MCP server.

Compoid MCP

Think of a digital repository organized this way as a highly structured "library of the future". Instead of just a pile of files, it’s a living ecosystem where data is categorized by Communities and described by rich Metadata, all made accessible to an AI through the Model Context Protocol (MCP). Compoid-mcp allows you to search and create records, download and upload artifacts.

AI Agents for Content Creation

For professionals, a repository like Compoid acts as an External Brain. By combining a massive, multi-modal library (docs, images, video) with an AI Agent via the compoid-mcp bridge, the LLM transforms from a simple chatbot into a sophisticated creative director that knows exactly what assets are "in the building."

Research Agent

The combination of Compoid’s structured repository and the compoid-mcp connection transforms the LLM into an Automated Research Associate. Instead of just retrieving documents, the AI synthesizes high-level insights from machine-learning-generated reports and converts them directly into professional-grade deliverables.

🚀 Quick Start

Configure your MCP client with the following settings to get started:

Claude Code If you are using the Claude Code terminal agent, run:

claude mcp add compoid --transport http https://mcpv.compoid.com/mcp --header "X-Compoid-Repo-Key: YOUR_API_KEY"

Manual Configuration 

{
  "mcpServers": {
    "Compoid": {
      "url": "https://mcpv.compoid.com/mcp",
      "disabled": false,
      "transportType": "streamable-http",
      "headers": {
        "X-Compoid-Repo-Key": "Your-Compoid-API-Key"
      }
    }
  }
}

Detailed instructions: https://www.compoid.com/documentation

Github repository: https://github.com/compoid/compoid-mcp.git


🤖 Multi-Agent Workflow

A streamlined approach to collaborative research and content creation using specialized AI agents


┌─────────────────────────────────────────────────────────────┐
│                    STAGE 1: RESEARCH                        │
│  Bot1: Search State → Research → Publish → Update State     │
│  Duration: 20-30 minutes                                    │
│  State: RESEARCH_COMPLETE: no → RESEARCH_COMPLETE: yes      │
└─────────────────────────────────────────────────────────────┘
                              │
                              ▼ (Bot2 polls state)
┌─────────────────────────────────────────────────────────────┐
│                   STAGE 2: EVALUATE                         │
│  Bot2: Poll State → Retrieve → Evaluate → Enhance → Update  │
│  Duration: 25-35 minutes                                    │
│  State: EVALUATION_COMPLETE: no → EVALUATION_COMPLETE: yes  │
└─────────────────────────────────────────────────────────────┘
                              │
                              ▼
┌─────────────────────────────────────────────────────────────┐
│                 STAGE 3: COMPLETE                           │
│  Both agents idle, state shows both complete                │
│  Total Duration: ~45-65 minutes                             │
│  Final output in ai-tutorials community                     │
└─────────────────────────────────────────────────────────────┘

🎯 Use Cases

Tools & Functions
For AI Agents: Empower your autonomous agents with direct access to Compoid's repository. AI agents can programmatically search, filter, and retrieve content using structured metadata, enabling them to perform complex research tasks and manage digital assets without human intervention.
  • Research Assistant: Search and download academic papers
  • Content Creator: Find and manage images/videos for projects
  • Data Analyst: Access datasets and quantitative analysis
  • Knowledge Manager: Organize and curate research collections
Documentation
For Humans: Interact with Compoid's vast repository using natural language commands and intuitive interfaces. Simply describe what you're looking for, and the system will help you discover, organize, and manage content across all platforms you use daily.
  • Natural Language Interface: "Find me papers about transformers"
  • Batch Operations: "Download all images from this community"
  • File Management: Download open-access records as zip archives and upload files via data URI
  • Cross-Platform: Use same tools in Cursor, Claude, VS Code, etc.

The Repository Structure

The repository isn't just a folder tree; it’s a hierarchy designed for both human navigation and machine readability.

Communities & Collection:

  • Communities: These serve as the top-level nodes (e.g., "Engineering," "Marketing," or "Legal"). They define the ownership and the high-level context of the data.
  • Collections: Within each community, data is further grouped into specific projects or document types. This allows for granular permissions and targeted searching.

Metadata: The Digital DNA:

Metadata is what allows an LLM to "understand" a file without necessarily reading the whole thing. It usually includes:

  • Descriptive: Title, author, keywords, and abstract.
  • Technical: File format, size, and checksums.
  • Administrative: License info, retention dates, and version history.

The Bridge: How the LLM Interacts with the Data

The Model Context Protocol (MCP) is the secret sauce. It acts as a standardized "translator" between the LLM and your repository. When you use an MCP server like compoid-mcp, you are giving the LLM a set of tools it can use to interact with your data. Instead of just "chatting," the AI can now execute specific functions inside the repository. Through the compoid-mcp connection, the LLM performs four primary actions:

  • Search Records: The LLM doesn't just do a keyword match. Because it understands the Metadata, it can perform complex queries.
  • Create Records: If you provide a raw document, the AI can summarize it, suggest relevant keywords (metadata), assign it to a Community, and call a function to create a new entry in the repository database.
  • Download Artifacts: The MCP connection allows the LLM to provide you with direct links or stream the content of a file. It identifies the correct "artifact" (the actual file) associated with a metadata record and retrieves it for your use.
  • Upload Artifacts: The LLM utilizes the compoid-mcp connection to programmatically upload files into specific communities while automatically generating and attaching the metadata necessary for instant searchability and organization.

The use of this website constitutes acceptance of our user agreement. Please read our privacy policy and content policy.