Content Analysis MCPs
Relevant source files
The following files were used as context for generating this wiki page:
Introduction
Within the AI Dev Atelier ecosystem, MCPs (Multi-Capability Providers) are specialized servers that extend the capabilities of AI agents by providing them with tools for complex content analysis tasks. These tasks include web search, academic research, documentation lookup, PDF extraction, and semantic code analysis. MCPs are defined in the mcp.json file and are central to skills like search and research, enabling agents to gather and process information from a wide variety of external and internal sources.
The integration of MCPs allows agents to perform deep analysis that goes beyond their native toolset. For example, an agent can use the openalex-research MCP to discover academic papers, the pdf-reader MCP to extract their content, and then synthesize the findings into an evidence card. This system is designed to be configurable, with API keys managed via a .env file and specific tool functionalities that can be filtered per agent.
Sources: AGENTS.md:3, AGENTS.md:50-65, README.md:38-50, skills/search/SKILL.md, skills/research/SKILL.md
Architecture and Configuration
MCPs function as a bridge between the AI agent and specialized data sources or processing engines. The agent invokes a tool provided by an MCP, which then handles the request, interacts with the external service, and returns the result. The configuration for these servers is centralized in mcp.json and loaded during the installation process.
Sources: AGENTS.md:12-19, README.md:21
This diagram illustrates the basic interaction flow between an AI agent and an MCP.
Sources: AGENTS.md:50-51, skills/search/SKILL.md:21
Core MCPs for Content Analysis
The repository defines several key MCPs dedicated to content analysis, each serving a distinct purpose.
| Server | Purpose | Required Key / Env Var | Associated Skill(s) |
|---|---|---|---|
tavily-remote-mcp | Web search and research | TAVILY_API_KEY | search |
context7 | Library documentation search | CONTEXT7_API_KEY (Optional) | search |
grep | GitHub code search | None | search (implied) |
openalex-research | Academic papers and citations | OPENALEX_EMAIL | research |
pdf-reader | PDF extraction | None | research |
paper-search | Multi-platform paper search | Optional | research |
zai-zread | GitHub repo semantic search (issues, PRs, docs) | Z_AI_API_KEY | search (implied) |
zai-vision | Image/video analysis, UI code gen, diagrams | Z_AI_API_KEY | ui-animation (implied) |
Sources: AGENTS.md:50-65, README.md:38-50, skills/search/SKILL.md:21-23, skills/research/SKILL.md:24-26
Configuration
MCPs are configured via mcp.json and require API keys to be set in a .env file before running install.sh.
Sources: README.md:19, AGENTS.md:67
Z.AI MCP Tool Filtering
For MCPs that provide multiple tools, such as zai-zread and zai-vision, it is possible to disable specific tools to save on API quota or avoid redundancy with an agent's native tools. This is configured in the agent's main configuration file (e.g., ~/.opencode/opencode.json) by setting the tool's key to false.
For example, the following configuration disables several Z.AI tools that have free, native equivalents.
{
"tools": {
"zai-zread_read_file": false,
"zai-zread_get_repo_structure": false,
"zai-vision_image_analysis": false,
"zai-vision_extract_text_from_screenshot": false
}
}
The documentation provides specific reasons for disabling these tools:
zai-zread_read_file: Can be replaced bywebfetchon the raw GitHub URL.zai-zread_get_repo_structure: Can be replaced by a directgh apicall.zai-vision_image_analysis/extract_text_from_screenshot: Can be replaced by the nativelook_attool.
Sources: AGENTS.md:69-87
MCPs in Action: The Research Skill
The research skill provides a clear example of how multiple MCPs are orchestrated to perform a complex content analysis workflow. The goal is to conduct academic research, synthesize findings into evidence cards, and generate a comprehensive report.
Sources: skills/research/SKILL.md:1, skills/research/references/guides/guide-workflow.md:3
Research Workflow
The following diagram shows how an agent uses the research skill and its associated MCPs to find and analyze academic papers.
This diagram synthesizes the workflow described in skills/research/references/guides/guide-workflow.md.
Sources: skills/research/references/guides/guide-workflow.md:31-70, skills/research/references/guides/guide-evidence-cards.md:9-25
Key MCPs in the Research Workflow
openalex-research: Used for the initial discovery phase to find highly cited foundational papers and recent works. Tools likeget_top_cited_worksandsearch_worksare employed.paper-search: Complements OpenAlex by searching other platforms like arXiv and PubMed.pdf-reader: Once a relevant paper is identified, this MCP'sread_pdftool is used to extract the full text content directly into the agent's context for analysis. This is a critical step for creating evidence cards.
Sources: skills/research/references/guides/guide-workflow.md:40-45, skills/research/references/guides/guide-evidence-cards.md:31-36, skills/research/SKILL.md:24-26
Summary
The Content Analysis MCPs are a cornerstone of the ai-dev-atelier project, providing the structured capabilities necessary for AI agents to interact with and analyze a vast range of content. By externalizing functionalities like web search, code analysis, and document extraction into configurable servers, the system allows agents to remain lightweight while having access to powerful, specialized tools. The research and search skills demonstrate how these MCPs are orchestrated to perform sophisticated, multi-step analysis tasks that are crucial for modern AI-assisted development.
Sources: AGENTS.md:4-5, README.md:17