MCP_Catalog
Verified Safeby Swissbit92
Overview
A local AI companion server that provides conversational chat with multiple personas, enhanced by RAG (Retrieval-Augmented Generation) using web search and MongoDB data, and featuring advanced memory management and emotional state tracking.
Installation
docker-compose up --buildEnvironment Variables
- OLLAMA_BASE
- PERSONA_MODEL
- PERSONA_TEMPERATURE
- MODEL_CONTEXT_WINDOW
- OLLAMA_TEMP_REWRITE
- OLLAMA_TEMP_SUMMARIZATION
- OLLAMA_TEMP_FACT_EXTRACTION
- BRAVE_API_KEY
- BRAVE_MAX_RESULTS
- BRAVE_SAFESEARCH
- BRAVE_SEARCH_TIMEOUT
- BRAVE_ENABLED_RARITIES
- MONGODB_ENABLED
- MONGODB_URI
- MONGODB_TIMEOUT
- MONGODB_MAX_RESPONSE_BYTES
- MONGODB_ENABLED_RARITIES
- MONGODB_CACHE_CURRENT_PRICE
- MONGODB_CACHE_TECHNICAL
- MONGODB_CACHE_HISTORICAL
- MONGODB_CACHE_TRADING
- MEMORY_EMBEDDING_MODEL
- MEMORY_SUMMARIZATION_INTERVAL
- MEMORY_FACT_EXTRACTION_INTERVAL
- PERSONA_DIR
- COORD_PORT
- COORD_URL
- COORDINATOR_DB_PATH
- BRAVE_MCP_IMAGE
- MONGODB_MCP_TRANSPORT
Security Notes
The system runs FastAPI with Ollama, Brave Search MCP (ephemeral Docker containers), and MongoDB MCP (long-running Docker container). Resource limits (memory, CPU, PIDs) are applied to spawned MCP containers, and MongoDB operations are explicitly restricted to read-only via a blacklist. CORS is configured for localhost, which is appropriate for development. CRITICAL NOTE: The `docker-compose.yml` mounts `/var/run/docker.sock` into the backend container, granting it root access to the Docker daemon. While this is a common pattern for local developer tools that manage other containers, it is a significant security risk in production or multi-tenant environments. The code attempts to mitigate risks by hardening spawned containers with limits and labels for orphan detection, but the underlying exposure of the Docker socket remains. For its stated 'local personal use' context, these mitigations are reasonable, but for broader deployment, this would need re-evaluation.
Similar Servers
osaurus
Osaurus is an AI edge runtime for macOS, enabling users to run local and cloud AI models, orchestrate tools via the Model Context Protocol (MCP), and power AI applications and workflows on Apple Silicon.
rag-server-mcp
Provides Retrieval Augmented Generation (RAG) capabilities to Model Context Protocol (MCP) clients by indexing project documents and retrieving relevant content for LLMs.
MCP-buddy
A local desktop or web application to manage and interact with multiple MCP (Model Context Protocol) servers, offering optional AI orchestration and enhancement for responses.
Little_MCP
A local AI assistant leveraging Retrieval-Augmented Generation (RAG) and multi-tool agents for document Q&A, real-time information, and SQL database interaction.