Adopt Model Context Protocol (MCP) for DANDI Agent Integration

[bendichter]

Background

Anthropic recently released the Model Context Protocol (MCP), an open-source protocol for connecting LLM agents with external tools. The protocol's design aligns well with our existing agent framework and presents an opportunity to standardize our approach.

Proposal

Implement MCP support for DANDI, enabling integration with Cline and other MCP-compatible agents, such as Claude Desktop. This would allow these agents to explore dandisets and DANDI assets similarly to the current neurosift chat functionality.

Benefits

1. Enhanced Development Workflow

   • Direct VS Code integration through Cline
   • Support for multi-file operations
   • Ability to propose changes to existing files
   • Seamless integration with standard scientific workflows

2. Improved Error Handling

   • Cline's built-in capability to detect, analyze, and automatically fix errors
   • Iterative debugging process

3. Extensible Tool Integration

   • Access to growing ecosystem of MCP servers
   • Potential integrations with other tools (e.g., GitHub issue search for debugging)
   • Standardized protocol for future tool development

Technical Considerations

• Adapt current framework to align with MCP specifications
• Develop DANDI MCP server implementation
• Ensure compatibility with existing neurosift chat functionality

Next Steps

1. Review MCP documentation and Cline integration examples
2. Assess required changes to current framework
3. Create proof of concept for DANDI MCP server
4. Plan migration strategy for existing functionality

Related Links

• Model Context Protocol (MCP)
• Cline GitHub Repository

[magland]

@bendichter This sounds like a good idea. I've had a lot of success using cline, so this seems like the way to go. I'll need to read up on how the protocol works.