Skip to content
← Back to posts
Macula Meets MCP: Giving AI Agents Structured Access to Creative Work
Making Content Accessible to AI: How Macula MCP Connects AI Agents to Creative Work by Daniel Maricic
cc-0 · Data mining: Allowed

Macula Meets MCP: Giving AI Agents Structured Access to Creative Work

  • macula
  • MCP
  • content discovery
  • AI agents
  • API design

Macula is a platform for photographers and creators to publish their work with proper licensing and metadata. We've been serving a public API (unified-link) for a while, but AI agents don't connect to REST APIs easily. We think they need MCP.

MCP (Model Context Protocol) is a standard way for AI agents to connect to external data sources. Think of it as a universal adapter for data access. A common language that lets agents query file systems or databases through a single service interface, without custom integration work for each one. When an AI agent connects to a service with MCP support, it gains access to a set of tools (actions it can perform), prompts (pre-built workflows), and resources (documentation and context).

We've built an MCP server for Macula. Here's what it does and why we think it matters.

How Macula's MCP Works

When an AI agent connects to Macula's MCP server, it establishes a connection to our public API. From there, it can query our database of published content, search by keywords to find relevant files, access user profiles and their public collections, get metadata including AI generation information and licensing details, and discover content through random exploration.

The 14 Tools

We've built 14 specialized tools organized by domain:

File Discovery

  • Get detailed information about any published file
  • Retrieve technical metadata (EXIF, XMP, IPTC)
  • List available renditions and presets
  • Find files by license type or AI usage permissions

User Exploration

  • Access creator profiles and their public directories
  • Browse paginated collections of someone's work
  • Discover random content for exploration or inspiration

Search & Browse

  • Search our keyword taxonomy
  • Find all files tagged with a specific keyword
  • Browse through user directory structures

The 14 Prompts

Beyond individual tools, we've built pre-configured prompts for common workflows:

Discovery Flows

  • discover_user_content: Explore a creator's full portfolio
  • random_exploration: Find inspiration through random content
  • search_discovery: Search by keywords to find relevant content

Analysis Flows

  • content_analysis: Analyze file metadata and quality
  • metadata_extraction: Extract structured data for processing

Rights & Licensing

  • license_discovery: Find content by license type
  • rights_audit: Audit usage rights across a collection
  • rights_verification: Verify specific rights for a use case
  • data_mining_discovery: Find content permitted for AI training

Optimization

  • rendition_optimization: Find the best version for a specific use (web, print, social)

Real-World Use Cases

Use Case 1: Content Discovery for a Project

An AI agent working on a blog post about sustainable architecture could search keywords for "sustainable" or "green building," examine specific candidates with get_file, filter by license with list_files_by_license, and find the right resolution for web display with get_file_presets.

Use Case 2: Building a Content Dataset

A researcher building a training dataset could use list_files_for_ai with allowed: "DMI-ALLOWED" to find permitted content, filter by specific licenses, extract technical specifications, and iterate through paginated results.

Use Case 3: Creator Research

An agent analyzing creative trends could access a creator's profile, explore their organization structure, analyze their output patterns, and find related creators.

Use Case 4: AI-Powered Photographer Portfolio

Sarah is a professional landscape photographer. She publishes her best work on Macula with clear licensing. Some images are CC-BY for maximum reach, others are All Rights Reserved for commercial licensing. She has AI data mining enabled on select images to allow AI model training while protecting her commercial work.

She connects Manus AI to Macula's MCP server, and here's what happens.

Morning: Reviewing Her Published Work

Sarah asks Manus to "Show me my recently published travel photos and their details." Manus uses get_user to access Sarah's profile and see her directories, uses list_user_files to see her published images, uses get_file to get detailed metadata for specific images, and presents her with a summary of her portfolio.

Midday: Preparing a Client Presentation

A client needs a photographer for a sustainable architecture magazine. Sarah asks Manus to "Prepare a portfolio of my architectural and nature photography, filtered by CC-BY license." Manus uses list_user_files with show: "images", uses list_files_by_license to filter by "Attribution (CC BY)", uses get_file_metadata to get technical specs for each image, and compiles a presentation-ready summary with image links, dimensions, and license info.

Afternoon: AI Agent Builds a Feature Page

A travel blog wants to feature Sarah's Iceland photography. The blog's AI agent (connected to Macula via MCP) uses list_files_by_keyword with keyword: "iceland", examines candidates with get_file, grabs the right image sizes with get_file_presets, and generates an article draft with properly attributed images, correct license links, and photographer credit automatically.

Sarah gets attribution and the blog gets content with proper licenses and credit baked in.

Managing AI Data Mining Permissions

Sarah wants to see which of her images are enabled for AI training. Manus uses list_files_for_ai with allowed: "DMI-ALLOWED" to see her AI-friendly images, uses list_files_for_ai with allowed: "DMI-UNSPECIFIED" to see images without a setting, and helps Sarah decide which additional images to enable.

Sarah doesn't need to manually update her portfolio across multiple platforms. Macula is her single source of truth. Images are hosted with full metadata and licensing. MCP access lets AI agents read her work with correct attribution. When she publishes new work, AI agents see it immediately. Every image has clear license and copyright info baked in.

Tools like Manus AI, Lovable, Cursor, and any other MCP-connected agent can now access her work properly. Not by scraping websites or guessing licensing, but through structured, permissioned access that respects her choices.

Security & Performance

Public by Design

All MCP-accessible content is public. No authentication required because the data is already meant to be accessible. This simplifies the architecture and removes the overhead of managing credentials for AI agents.

Rate Limiting

We use two-layer rate limiting to ensure fair access:

  • Slow-down layer: Progressive delays after 100 requests prevent abuse
  • Hard limit: 200 requests per minute maximum

Input Validation

Every request is validated and sanitized. String inputs are checked against strict patterns, length limits prevent oversized requests, and dangerous characters get stripped.

Developer Experience

For developers building AI-powered applications:

// Example: Finding CC-licensed images
const response = await mcpClient.callTool('list_files_by_license', {
  license: 'Attribution (CC BY)',
  limit: 10,
  page: 0,
});

// Example: Getting file metadata
const fileData = await mcpClient.callTool('get_file', {
  unifiedId: 'abc123xyz',
});

The MCP interface abstracts away our internal implementation. You don't need to understand our database schema, API versioning, or caching strategy. The tools are designed to be intuitive and self-documenting.

Complete Reference

All 14 Tools

File Tools (6)

Tool Description
get_file Get file information by unifiedId. Includes title, description, creator, links, assets, presets, size, copyright info, AI info
get_file_metadata Get full EXIF/XMP/IPTC metadata. Optional a parameter for specific metadata fields
get_file_presets Get available renditions (sys_sm, sys_lg, open_graph, etc.) with size and MIME info
get_file_json_schema Get the JSON Schema for the get_file tool output
get_metadata_json_schema Get the JSON Schema for the metadata tool output
list_files_for_ai List files filtered by data mining allowance (DMI-ALLOWED or DMI-UNSPECIFIED)

User Tools (3)

Tool Description
get_user Get user profile. Includes name, bio, directories, stats
list_user_files List user's files (paginated). Filter by type: images, videos, audio, documents
list_random_files Get random files for discovery and inspiration

Directory Tools (2)

Tool Description
get_directory Get directory metadata by nickname and pathCid
get_directory_files Get paginated files from a directory

Search Tools (3)

Tool Description
search_keywords Search the keyword taxonomy
list_files_by_keyword Get all files tagged with a specific keyword
list_files_by_license List files by license type (CC BY, All Rights Reserved, etc.)

All 14 Prompts

Prompt Description
discover_user_content Explore a creator's full portfolio and content library
content_analysis Analyze file metadata, quality, and technical specifications
search_discovery Search for content using keywords and filters
license_discovery Find content by license type (CC BY, CC BY-SA, All Rights Reserved, etc.)
content_curation Curate content collections based on themes or criteria
rights_audit Audit usage rights and permissions across a collection
rights_verification Verify specific rights for a particular use case
rendition_optimization Find the optimal file version for web, social, print, or API use
creator_ecosystem Explore a creator's profile, directories, and related creators
metadata_extraction Extract and analyze file metadata in structured format
random_exploration Discover content through random exploration
directory_deep_dive Explore user directory structures and organization
service_info Get information about the Macula service and its capabilities
data_mining_discovery Find content permitted for AI training and data mining

Resources (1)

Resource Description
instructions Service documentation and usage guidelines for AI agents

All tools are read-only. You can query and analyze, but not modify data. This keeps the system safe and predictable.

Works With

Any AI agent or platform that supports MCP:

  • Manus AI. Full AI agent for complex workflows.
  • Lovable. Build apps with AI assistance.
  • Cursor. AI-powered code editor.
  • Claude Desktop. Anthropic's MCP integration.
  • Custom agents. Build your own with the SDK.

The Ecosystem Advantage

When photographers host on Macula, their work becomes part of a growing ecosystem:

  • For creators: One place to publish, with licensing and copyright built-in
  • For AI agents: Standardized access to millions of files with correct attribution
  • For everyone: Better licensing compliance, less copyright confusion, more fair use of creative work

Getting Started

  1. Connect to our MCP server at https://u.macula.link/mcp
  2. Explore available tools. The server will describe what it can do.
  3. Try a prompt. Start with discover_user_content or random_exploration.
  4. Build your workflow. Chain tools together for complex tasks.

As AI agents get more capable, being able to discover and work with creative content matters more. MCP gives us the interface to make that happen.

We're continuing to expand our toolset based on real usage patterns. If you're building AI-powered content applications, we'd love to hear what you'd like to see.

For technical details on our MCP implementation, see Building a Public MCP Server: From Zero to Production.

Published