snowflake-mcp-lambda

snowflake-mcp-lambda

A remote MCP Server for Snowflake. Deployed as a AWS Lambda Function

🚀 Developer Setup (REQUIRED)

IMPORTANT: Our pre-commit hooks exactly mirror CI. This means longer commit times but zero CI failures.

# One-time setup (takes ~2 minutes)
./scripts/setup-dev.sh

# That's it! All commits will now run the same checks as CI:
# - Ruff linting & formatting
# - MyPy type checking
# - Full pytest suite with 85% coverage requirement
# - Security scanning

Why Full CI Parity in Pre-commit?

1. No CI Debugging: Fix issues locally, not through CI logs
2. Resource Efficiency: Don't waste CI minutes on preventable failures
3. Guaranteed Success: If it commits locally, it passes CI
4. Better DX: Immediate feedback, no context switching

Yes, commits take 30-60 seconds. But that's better than 5-10 minute CI debug cycles.

🤖 Why This Matters for AI Code Assistants

When working with AI code assistants (like Claude), comprehensive pre-commit hooks are essential:

The Problem

AI assistants can't directly see CI outputs. When CI fails, the human developer must:

1. Navigate to CI logs
2. Copy error messages
3. Paste them back to the AI
4. Wait for a fix
5. Push again
6. Repeat until green ✅

This creates a frustrating human-in-the-middle debugging loop that wastes everyone's time.

The Solution

With full CI parity in pre-commit hooks:

• AI assistants get immediate feedback on their code
• Errors are caught before pushing
• The AI can fix issues in the same conversation
• No context switching or copy-pasting required

Best Practices for AI-Assisted Development

1. Make pre-commit hooks comprehensive - Include all CI checks
2. Fail fast, fail locally - Better to wait 60 seconds than debug through logs
3. Clear error messages - AI assistants can parse and fix clear errors
4. No surprises - What passes locally MUST pass in CI

This approach transforms AI code assistants from "helpful but sometimes frustrating" to "genuinely reliable development partners."