Skip to content

PLAN.md

Latest commit

 

History

History
76 lines (62 loc) · 3.69 KB

PLAN.md

File metadata and controls

76 lines (62 loc) · 3.69 KB

Oxigraph MCP Tools — Project Plan

Vision

Expose an Oxigraph RDF triplestore as MCP (Model Context Protocol) tools for Claude Code. A single Rust MCP server provides generic RDF/SPARQL tools alongside language-specific code-loading tools that parse source code into an RDF knowledge graph — enabling an LLM coding agent to query and reason about a project's structure, dependencies, and relationships.

Milestones

M0 — Project Foundation ✅

  • Define project goals, architecture, and specifications
  • Write PLAN.md, TASKS.md, SPECIFICATIONS.md
  • Set up repository structure and .gitignore

M1 — Core Server and Generic RDF Tools ✅

Minimal working MCP server with the 4 generic RDF tools.

  • Set up Rust project with Cargo.toml (oxigraph, rmcp, tokio, serde, schemars)
  • Implement MCP server entry point with stdio transport
  • Implement Oxigraph store initialization (on-disk RocksDB, configurable path)
  • Implement sparql_query tool
  • Implement sparql_update tool
  • Implement load_rdf tool (file path and inline content)
  • Implement list_graphs tool
  • Manual testing with Claude Code

M2 — LanguageLoader Trait and Rust Loader ✅

Plugin system foundation plus the first code loader.

  • Define the LanguageLoader trait
  • Implement loader registry and auto-detection
  • Implement load_code generic dispatcher tool
  • Implement Rust loader (load_rust_code):
    • Cargo.toml parsing (project metadata, dependencies)
    • .rs file AST extraction (modules, functions, structs, enums, traits, impls, use statements)
    • Module hierarchy resolution
  • Manual testing: load a Rust project and query its structure via SPARQL

M3 — Python Loader

  • Implement Python loader (load_python_code):
    • pyproject.toml / setup.py / requirements.txt parsing
    • .py file AST extraction (modules, functions, classes, decorators, imports, docstrings)
  • Manual testing: load a Python project and query its structure via SPARQL

M4 — TypeScript Loader ✅

  • Implement TypeScript loader (load_ts_code):
    • package.json parsing (metadata, dependencies)
    • .ts/.tsx/.js/.jsx file AST extraction via oxc_parser (modules, functions, classes, interfaces, type aliases, enums, imports/exports, JSDoc)
  • Manual testing: load a TypeScript project and query its structure via SPARQL

M5 — Testing and Documentation

  • Integration tests for generic RDF tools
  • Integration tests for each language loader
  • User-facing README with installation and usage instructions
  • Claude Code MCP configuration examples

M6 — Git History Loader

Load git commit history into the RDF knowledge graph, enabling queries that join code structure with change history.

  • Add git2 crate (libgit2 bindings) for native repository access
  • Implement git history loader in loaders/git.rs:
    • Walk commit graph from HEAD or specified branch
    • Extract commit metadata (hash, author, committer, date, message, parents)
    • Extract per-commit file diffs (added, modified, deleted, renamed)
    • Generate code:Commit and code:FileChange RDF triples
    • Cross-graph linking: code:FileChange → code:Module via code:affectsModule
  • Implement load_git_history tool in tools/git.rs
  • All triples written to the default graph (unified with code loaders)
  • Unit tests with temporary git repositories
  • Manual testing: load history and run SPARQL queries joining git history with code structure

M7 — Advanced Features

  • Named graph management tools (drop_graph, export_rdf)
  • Store statistics / introspection tool (store_stats)
  • Namespace prefix management (list_namespaces, add_namespace)
  • Bulk loading with progress reporting
  • Additional language loaders (Go, Java, C/C++, etc.)