Out of Context
An MCP server for managing contextual data as markdown files with metadata, enabling agents to save, retrieve, search, and delete contexts using simple CRUD operations.
README
Out of Context
An MCP (Model Context Protocol) server for managing context using simple CRUD operations. Stores contexts as markdown files (.mdc) with YAML frontmatter, allowing agents to save, retrieve, search, and manage context by name.
Features
- Simple CRUD Operations: 5 basic tools for context management (put, list, get, search, delete)
- Markdown Storage: Contexts stored as .mdc files (markdown with YAML frontmatter)
- Agent-Recognizable Names: Use meaningful names instead of UUIDs
- Bulk Operations: Support for bulk put, get, and delete operations with robust parameter handling
- Pydantic Validation: Type-safe parameter validation with automatic schema generation
- Text Search: Search contexts by query string across metadata and content
- MCP Integration: Works with any MCP-compatible platform (Cursor, Claude Desktop, etc.)
Quick Start
Installation
pip install hjeon139-mcp-outofcontext
MCP Server Configuration
Add to your MCP platform configuration (e.g., Cursor or Claude Desktop):
{
"mcpServers": {
"out-of-context": {
"command": "hjeon139_mcp_outofcontext",
"env": {
"OUT_OF_CONTEXT_STORAGE_PATH": "out_of_context"
}
}
}
}
Verify Installation
In your MCP platform, check that tools like put_context, list_context, get_context, search_context, and delete_context are available.
Usage Examples
Add Context
Single operation:
{
"tool": "put_context",
"arguments": {
"name": "api-design-notes",
"text": "# API Design Notes\n\nKey decisions about the REST API...",
"metadata": {
"type": "note",
"tags": ["api", "design"]
}
}
}
Bulk operation:
{
"tool": "put_context",
"arguments": {
"contexts": [
{
"name": "context-1",
"text": "First context",
"metadata": {"type": "note"}
},
{
"name": "context-2",
"text": "Second context"
}
]
}
}
List Contexts
{
"tool": "list_context",
"arguments": {
"limit": 10
}
}
Returns list of contexts sorted by creation date (newest first).
Get Context
Single operation:
{
"tool": "get_context",
"arguments": {
"name": "api-design-notes"
}
}
Bulk operation:
{
"tool": "get_context",
"arguments": {
"names": ["context-1", "context-2", "context-3"]
}
}
Search Contexts
{
"tool": "search_context",
"arguments": {
"query": "API design",
"limit": 5
}
}
Searches in both YAML frontmatter (metadata) and markdown body (text content).
Delete Context
Single operation:
{
"tool": "delete_context",
"arguments": {
"name": "old-context"
}
}
Bulk operation:
{
"tool": "delete_context",
"arguments": {
"names": ["context-1", "context-2"]
}
}
Storage Format
Contexts are stored as .mdc files (markdown with YAML frontmatter) in the out_of_context/contexts/ directory.
File format:
---
name: api-design-notes
created_at: 2025-12-14T12:51:27.123456
type: note
tags: [api, design]
---
# API Design Notes
Key decisions about the REST API design...
- Use RESTful conventions
- Version in URL path
Name requirements:
- Filename-safe: alphanumeric characters, hyphens, and underscores only
- Unique: overwriting an existing name replaces the old context (with warning)
Documentation
- Installation Guide - Setup and configuration
- Usage Guide - Detailed usage instructions
- Development Guide - Development setup and contribution guidelines
- API Documentation - Complete tool reference
Key Concepts
- Context: A markdown document with YAML frontmatter (metadata) and markdown body (content)
- Name: Agent-recognizable identifier (e.g., "api-design-notes", "bug-fix-context")
- Storage: Individual .mdc files in
out_of_context/contexts/directory - Bulk Operations: Process multiple contexts in a single call (put, get, delete)
Architecture
The server provides a simple file-based storage system built with FastMCP:
Key Components:
- FastMCP Server: Modern MCP server implementation with middleware support
- MDCStorage: Manages .mdc file operations (save, load, list, search, delete)
- CRUD Tools: 5 tool handlers using
@mcp.tool()decorators for automatic registration - AppStateMiddleware: Dependency injection pattern for clean state management
Storage:
- Each context is one .mdc file
- YAML frontmatter for metadata
- Markdown body for content
- Simple text-based search
Development
Setup
# Clone repository
git clone <repository-url>
cd out_of_context
# Create environment
hatch env create
# Install dependencies
hatch run update-deps
Run Tests
# Unit tests
hatch run pytest -m 'unit'
# Integration tests
hatch run pytest -m 'integration'
Code Quality
# Lint and format
hatch run lint-fix
hatch run fmt-fix
# Type check
hatch run typecheck
# Full release pipeline
hatch run release
See Development Guide for detailed setup and contribution guidelines.
Project Structure
out_of_context/
├── src/hjeon139_mcp_outofcontext/ # Main package
│ ├── fastmcp_server.py # FastMCP instance + middleware
│ ├── main.py # Entry point
│ ├── tools/
│ │ ├── crud/ # CRUD operations (put, get, delete)
│ │ └── query/ # Query operations (list, search)
│ ├── storage/ # MDC storage layer
│ ├── app_state.py # Application state
│ ├── config.py # Configuration
│ └── prompts.py # MCP prompts
├── tests/ # Test files (195 tests)
├── docs/ # Documentation
└── pyproject.toml # Project configuration
Contributing
Contributions welcome! Please:
- Follow Conventional Commits format
- Add tests for new functionality
- Update documentation as needed
- Run pre-commit checklist before submitting
See Development Guide for detailed contribution guidelines.
License
Apache 2.0 - See LICENSE file for details.
References
- MCP Protocol: Model Context Protocol
- Steering Documentation: docs/steering/ - Development guidelines
Status
Version: 1.0.0 (Launch Release)
Status: Production Ready
Features:
- ✅ Basic CRUD operations (put, list, get, search, delete)
- ✅ Markdown file storage (.mdc format)
- ✅ Agent-recognizable names
- ✅ Bulk operations support with robust parameter handling
- ✅ Pydantic validation for type-safe parameters
- ✅ Automatic JSON schema generation for MCP clients
- ✅ Text search across metadata and content
- ✅ Built with FastMCP for improved developer experience
Support
- Documentation: See docs/ directory
- Issues: Open an issue on GitHub
- Questions: Use GitHub Discussions
Acknowledgments
Built with:
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
Neon Database
MCP server for interacting with Neon Management API and databases
E2B
Using MCP to run code via e2b.
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.