MCPedia

MCPedia is a knowledge base server that stores, searches, and manages reusable knowledge entries for AI agents via the MCP protocol. It acts as a persistent memory store for skills, rules, patterns, guides, and references, enabling agents like Cursor, Codex, Claude, and other MCP-enabled tools to read and write knowledge across sessions. You can save entries that codify coding patterns, project conventions, or debugging tricks, then search or filter entries by language, domain, project, or tags. The same knowledge base can be shared across different agents, so you won’t lose context when switching tools. MCPedia also supports a write lock to temporarily prevent modifications, giving you control over when agents can alter the knowledge base. Tools exposed by the API include search_entries for full-text search, get_entry to retrieve a single entry with full content, get_entries_by_context to fetch contextual results with content, and listing operations to browse metadata and tags. Use the MCP endpoint POST /mcp to interact with these tools from your agents.

To get started, run the server using the MCPedia command, then send requests such as searching for a topic with a language filter, retrieving a specific entry by its slug, or listing entries to discover relevant knowledge to inject into agent context.

Prerequisites:

• Node.js (LTS version recommended)
• npm or corepack

Installation steps:

# 1) Install the MCPedia package via npx (or install a binary if provided by the project)
# This will pull the MCPedia server package and install as needed
npx -y pouriya-mcpedia

# 2) (Optional) Install as a local package if you plan to run from source
# git clone <repository-url> && npm install

Usage:

# Start the MCPedia server (if the package provides a start script via npx or npm)
npx -y pouriya-mcpedia

# If you installed locally
cd pours/mcpedia
npm run start

Configuration:

• Review environment variables and defaults in the documentation or README that accompany the server package.
• If a write lock token is used, set the appropriate environment variable or configuration file as documented by the project.

Notes & tips:

• The MCP API endpoints include search_entries, get_entry, get_entries_by_context, list_entries, and list_tags. Use these to build richer agent prompts and memory access patterns.
• If you plan to run multiple agents against the same knowledge base, consider enabling a write lock during maintenance windows to prevent concurrent modifications.
• The knowledge base supports tags for cross-cutting discovery; ensure you tag entries appropriately to improve retrieval.
• When sharing knowledge across agents, keep sensitive or project-private information in restricted entries and manage access through proper authentication and locking mechanisms.
• If you encounter performance issues with large knowledge bases, evaluate indexing options for the full-text search and review the hardware requirements for the underlying database.