MCP PostgreSQL Server
# MCP Server Test Collection
[](https://opensource.org/licenses/MIT)
[](https://github.com/OpenBrokerRemover)
[](https://github.com/OpenBrokerRemover/pulls)
This repository contains a collection of tests and implementations for various Model-Controller-Provider (MCP) server architectures. Each implementation focuses on different aspects of MCP pattern and database interactions.
## Current Implementations
### 1. MCP PostgreSQL Server (`mcp-psql/`)
A PostgreSQL-specific implementation that:
- Provides safe database access through a structured API
- Enables AI systems to interact with databases
- Implements read-only query validation
- Includes pre-built analysis templates
- Supports schema exploration and data analysis
### 2. MCP Figma Server (`mcp-figma/`)
A Figma-specific implementation that:
- Enables AI systems to interact with Figma designs
- Provides structured access to Figma files and components
- Implements design analysis and manipulation capabilities
- Supports AI-driven design suggestions and modifications
- Includes pre-built design templates and patterns
## Purpose
This repository serves as:
- A testing ground for different MCP implementations
- A reference for MCP pattern best practices
- A collection of database and design tool interaction patterns
- A showcase of AI integration approaches with various systems
## Getting Started
Each implementation in this collection has its own setup instructions and documentation. Please refer to the specific implementation's README for details.
## Contributing
New MCP implementations and test cases are welcome! Please follow these guidelines:
1. Create a new directory for your implementation
2. Include comprehensive documentation
3. Follow the existing project structure
4. Add appropriate tests
## License
MIT
TDQS
Scored across 6 tools
Each tool has a clearly distinct purpose with no ambiguity. 'execute' handles INSERT/UPDATE/DELETE, 'query' handles SELECT, 'list_tables' and 'list_schemas' are separate listing functions, 'describe_table' provides metadata, and 'connect_db' is for connection management. The separation between 'execute' and 'query' is particularly well-defined.
Most tools follow a consistent verb_noun pattern (describe_table, list_schemas, list_tables). The main deviation is 'execute' and 'query' which use single verbs without objects, and 'connect_db' uses a verb_object pattern but with abbreviated 'db' instead of 'database'. The naming is still readable and mostly predictable.
Six tools is well-scoped for a PostgreSQL server. This covers essential database operations: connection management, schema/table discovery, metadata inspection, and both read (SELECT) and write (INSERT/UPDATE/DELETE) operations. Each tool earns its place without bloat.
The toolset covers core PostgreSQL workflows well: discovery (list_schemas, list_tables), inspection (describe_table), and query execution (query, execute). Minor gaps include lack of transaction control tools (BEGIN, COMMIT, ROLLBACK) and database management operations (CREATE/DROP database/table), but agents can work around these using the existing execute tool.