SDL

概要

SDL-MCP is a Model Context Protocol (MCP) server that transforms codebases into a queryable, versioned knowledge system for AI agents. Rather than having agents read entire files, it indexes repositories into a SQLite-backed symbol ledger, extracting symbols and their dependency relationships. This foundation enables agents to search, retrieve precise context via 'symbol cards', build focused graph slices for tasks, and refresh context incrementally. The system also supports responsible code access through policy-gated options, offering skeleton views, hot-path excerpts, and controlled full-code windows. Ultimately, SDL-MCP facilitates faster, cheaper, and safer agent assistance by reducing token usage, enhancing context relevance, and providing better governance through auditing.

主な機能

• Multi-language repository indexing with Tree-sitter adapters
• Symbol cards with signatures, dependencies, metrics, and versioning
• Graph slices for building task-specific, token-budgeted context
• Delta analysis and blast radius support for version changes and PRs
• Policy-gated code access ladder (skeleton, hot-path, and full window)
• 1 GitHub stars

ユースケース

• Optimizing AI agent context retrieval to reduce token usage and improve relevance
• Enhancing agent development workflows with dependency-aware code understanding
• Improving code review and impact analysis through delta and PR risk analysis