
U2 MCP Server
io.github.bpamiriPublished by GitHub account bpamiriv0.1.1Updated Oct 27, 2026
Connect AI assistants to Rocket Universe/UniData MultiValue databases via MCP.
Overview
Connects an AI assistant to Rocket Universe/UniData MultiValue databases so it can query, read and modify records in plain language.
- What you can do
- Connect, disconnect and list connections to Universe/UniData servers
- List files, read single or multiple records, write and delete records, get file statistics
- Run RetrieVe/UniQuery statements, TCL/ECL commands and SELECT lists
- Browse dictionaries: list items, get field definitions, describe files
- Call cataloged BASIC subroutines and manage transactions with begin, commit and rollback
- Save, list, search and delete notes about the database that persist across sessions
- When to use it
- Useful when an assistant needs to work with a Pick/MultiValue database — exploring file structures, answering questions about records, or running queries — without hand-writing RetrieVe syntax. Also suited to teams that want a shared, centrally deployed connection with saved schema knowledge.
- Requirements
- Local Python 3.10+ process, run via pip or uvx; the uopy package installs as a dependency. Needs network access to a Universe or UniData server and the environment variables U2_HOST, U2_USER, U2_PASSWORD and U2_ACCOUNT, plus optional U2_SERVICE, U2_PORT, U2_SSL and related settings. An HTTP/SSE mode and OAuth-based deployment are also documented.
Installation
In SourceWeft
- Open U2 MCP Server in the dashboard and add it to a workspace.
- Enable the server for the chats that should use its tools.
Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.
Other MCP clients
Follow the launch instructions in the repository.
README
u2-mcp
[PyPI version] [License] [Python] [MCP]
Connect Claude to your Universe/UniData database using natural language.
u2-mcp is an MCP server that lets AI assistants like Claude query and interact with Rocket Universe and UniData MultiValue databases. Ask questions in plain English and get real answers from your data.
The first MCP server for the Pick/MultiValue database family.
What Can You Do With This?
Instead of writing RetrieVe queries, just ask Claude:
How It Works
- You ask Claude a question about your database
- Claude calls u2-mcp with the appropriate tool
- u2-mcp queries your Universe/UniData server
- Results come back through Claude in a readable format
New to MCP? See What is MCP? for a complete explanation.
Documentation
Features
- Connection Management - Connect to Universe/UniData servers with connection pooling and SSL support
- File Operations - Read, write, and delete records while preserving MultiValue semantics
- Query Execution - Run RetrieVe/UniQuery statements with safety controls
- Dictionary Access - Explore file structures and field definitions
- BASIC Subroutine Calls - Execute cataloged BASIC programs
- Transaction Support - Full transaction management with commit/rollback
- AI-Optimized - Built-in query examples and syntax help for better LLM interactions
Installation
Or using uvx for isolated execution:
Quick Start
Prerequisites
- Python 3.10+
- Access to a Rocket Universe or UniData server
- The
uopypackage (installed automatically as a dependency)
Configuration
Set environment variables for your database connection:
Claude Desktop Configuration
Add to your claude_desktop_config.json:
Usage Examples
Once connected, you can use natural language to interact with your Universe database:
- "List all customers in California"
- "Read record CUST001 from the CUSTOMERS file"
- "Show me the dictionary for the ORDERS file"
- "Count all open invoices over $1000"
Available Tools
Connection Management
connect- Establish connection to Universe/UniData serverdisconnect- Close connectionslist_connections- Show active connections
File Operations
list_files- List available files in the accountread_record- Read a single record by IDread_records- Read multiple recordswrite_record- Write/update a recorddelete_record- Delete a recordget_file_info- Get file statistics
Query Execution
execute_query- Run RetrieVe/UniQuery statementsexecute_tcl- Run TCL/ECL commandsget_select_list- Execute SELECT and return record IDs
Dictionary Access
list_dictionary- List dictionary itemsget_field_definition- Get field detailsdescribe_file- High-level file description
Advanced Features
call_subroutine- Call BASIC subroutinesbegin_transaction/commit_transaction/rollback_transaction- Transaction management
Knowledge Persistence
save_knowledge- Save learned information about the databaselist_knowledge- List all saved knowledge topicsget_knowledge_topic- Retrieve specific topicsearch_knowledge- Search across saved knowledgedelete_knowledge- Remove a knowledge topic
Configuration Options
Knowledge Persistence
The MCP server includes a knowledge persistence feature that allows Claude to save and recall learned information about your database across sessions. This eliminates repetitive discovery work and speeds up future interactions.
How It Works
When Claude discovers useful information about your database (file purposes, field meanings, working queries), it can save this knowledge using the save_knowledge tool. This information is stored in a markdown file and automatically available in future conversations via the u2://knowledge resource.
What Gets Saved
- File descriptions - What each file contains (e.g., "AR-CUST is the customer master file")
- Field mappings - Which fields contain what data (e.g., "Field 1 is customer name")
- Query patterns - Queries that produced good results
- Data formats - Date formats, code meanings, conversion notes
- Relationships - How files relate to each other
Storage Location
Knowledge is stored in ~/.u2-mcp/knowledge.md by default. You can customize this location using the U2_KNOWLEDGE_PATH environment variable:
Example Usage
After Claude discovers that AR-CUST contains customer records:
In the next conversation, Claude will already know this and won't need to rediscover it.
Centralized Deployment (HTTP/SSE Mode)
For team environments, you can run u2-mcp as a centralized HTTP server instead of local stdio mode. This allows:
- Single database connection for the whole team
- Shared knowledge base
- Centralized credential management
- Access from any MCP-compatible client
Running the HTTP Server
Docker Deployment
HTTP Configuration Options
Connecting Clients
The HTTP server exposes an SSE (Server-Sent Events) endpoint at /sse. MCP clients that support remote servers can connect to:
Note: As of late 2024, Claude Desktop only supports local stdio MCP servers. HTTP/SSE mode is for:
- Custom MCP clients
- Future Claude Desktop versions with remote server support
- Integration with other AI tools that support MCP over HTTP
Claude.ai Integration (Streamable HTTP + OAuth)
Deploy u2-mcp as a Custom Connector in Claude.ai with enterprise authentication via your identity provider (Duo, Auth0, Okta, etc.).
Quick Setup
- Deploy u2-mcp with OAuth enabled:
- Configure OAuth in your
.env:
- Add to Claude.ai: Go to Settings → Connectors → Add Custom Connector → Enter your URL
Users authenticate through your IdP and can then query your database directly from Claude.ai.
Supported Identity Providers
- Cisco Duo - Enterprise MFA and SSO
- Auth0 - Flexible identity platform
- Any OIDC Provider - Okta, Azure AD, Google Workspace, Keycloak, etc.
See the OAuth Guide for detailed setup instructions.
Security
- Credentials are never logged
- Configurable command blocklist prevents dangerous operations
- Optional read-only mode for safe exploration
- Honors Universe's built-in security model
See SECURITY.md for reporting security vulnerabilities.
Contributing
We welcome contributions! Please see CONTRIBUTING.md for guidelines.
Documentation
Getting Started
- Quickstart Guide - Get running in 10 minutes
- What is MCP? - Understanding MCP and how u2-mcp works
- Installation Guide - Detailed installation instructions
Reference
- Configuration Reference - All environment variables
- Tools Reference - Complete tool documentation
- Usage Examples - Common usage patterns
- OAuth & Claude.ai Integration - Deploy as a Claude.ai Custom Connector
License
This project is licensed under the Apache License 2.0 - see the LICENSE file for details.
Acknowledgments
- Rocket Software for the
uopypackage - Anthropic for the Model Context Protocol specification
- The MultiValue database community
Source: README.md at commit f427768
Tools
0Version history
1- v0.1.1LatestSep 16, 2026
Score
50% of the applicable checks met
How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.
- Reached in the last 30 days25 pts
Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.
- Auth declared10 pts
Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.
- Tools listed15 pts
Answer tools/list without a credential, or ship a package that starts in a sandbox without one.
- DescriptionRequired15 pts
- Logo10 pts
- Documentation15 pts
- License10 pts