GlitterKill/sdl-mcp

Symbol Delta Ledger (SDL-MCP) is a policy-centered context budget layer for coding agents: Symbol-graph intelligence combined with precision tools. It turns sprawling codebases into compact, high-signal context that saves tokens, speeds up workflows, and improves agent output.

What it solves

SDL-MCP addresses the "context budget" problem for AI coding agents. Instead of feeding entire source files into an LLM—which consumes tokens and adds noise—it provides a structured way for agents to discover and retrieve only the specific pieces of code and metadata necessary for a given task.

How it works

The system indexes a codebase into a symbol graph. It uses a tiered retrieval approach called the "Iris Gate Ladder," allowing agents to escalate from compact metadata (symbol cards) to skeletons, hot paths, and finally bounded source code windows. It leverages tree-sitter for structure and can integrate SCIP or language-provider facts for precise cross-referencing. The project implements the Model Context Protocol (MCP) to communicate with AI clients via stdio or HTTP.

Who it’s for

It is designed for developers building or using AI coding agents that need to navigate large repositories efficiently without overwhelming the model's context window.

Highlights

  • Symbol Cards: Compact metadata for each symbol, allowing agents to start exploration without opening full files.
  • Graph Slicing: Retrieves subgraphs based on repository relationships and task scope rather than simple directory boundaries.
  • Delta Packs: Analyzes changes between versions to identify affected symbols and calculate the "blast radius" for PR reviews.
  • Live Indexing: Supports draft-buffer updates via a live overlay for real-time context updates before files are saved.
  • Development Memories: Opt-in storage for decisions and task notes linked directly to the repository.
  • Tool Gateway: Offers flexible tool surfaces (Flat, Gateway, or Code Mode) to optimize the number of tools exposed to the AI client.

Related

  • Project
    Context-Engine-AI/Context-Engine

    A semantic code search and symbol intelligence system that provides AI coding assistants with deep codebase context, memory, and navigation tools via MCP.

  • Project
    forloopcodes/contextplus

    An MCP server that turns large codebases into searchable, hierarchical feature graphs using RAG, AST parsing, and spectral clustering for high-accuracy AI coding context.

  • Project
    CodeGraphContext/CodeGraphContext

    A toolkit and MCP server that indexes local code repositories into a graph database to provide AI agents and developers with structural context like call chains and class hierarchies.

  • Dispatch
    Code Execution with MCP: Building More Efficient AI Agents

    Anthropic introduces a method for AI agents to interact with Model Context Protocol (MCP) servers via code execution rather than direct tool calls, reducing token consumption and improving efficiency.

  • Project
    trailhq/Graft

    Graft is an open‑source TypeScript/Node.js tool that builds a local, markdown‑based knowledge graph of a codebase (summaries, key snippets, typed links) using tree‑sitter for structural parsing and optional LLM summarisation. The graph is cached in a `graft/` folder, refreshed only on file changes, and automatically wired into coding agents like Claude Code, Cursor, Codex, and Gemini. Benchmarks claim up to 46 % fewer tool calls, 42 % token savings, 60 % latency reduction, and a 12‑point correctness boost on SWE‑bench. Integration is via a single `graft init` command that writes agent‑specific instruction files; the CLI also offers search, visualization, and telemetry controls. The project is MIT‑licensed and targets faster, cheaper AI‑assisted development.