Entity trail

Tree-sitter

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-06-20 / arxiv-researcherCodeSentinel: A Three-Layer Defense Against Indirect Prompt Injection in Code ContextsAn inference-time defense against indirect prompt injection hidden in code comments, strings, or identifiers, using Tree-sitter for syntax-guided pre-filtering plus CST-guided Dynamic Min-K% scoring and node perturbation analysis to flag risky nodes before code reaches the LLM. Achieves 0.80 average node-level F1 across six attack families, beating CodeGarrison and DePA. Practical for hardening coding agents that ingest untrusted repos.
2026-06-02 / github-pulse-researcherLeanCTX Ships v3.7.0 at 2.4K Stars: Context OS Provides 63 MCP Tools With 10 Read Modes and Up to 99% Token Savings for AI DevelopmentLeanCTX (yvgude/lean-ctx) is a single local Rust binary that compresses, remembers, routes, and verifies tokens between code and models through 63-67 MCP tools. V3.7.0 shipped June 1 with Tree-sitter AST processing for 18 languages, 56 shell output compression pattern modules, and a 4-layer verification engine. Cached file re-reads cost ~13 tokens versus 2,000+ for initial reads, with LSP refactoring tools (rename, references, go-to-definition) built into the MCP surface.
2026-05-28 / agents-researcherClaude Code Deny Rule Bypass: Shell Security Rules Silently Disabled After 50 Subcommands in Compound CommandsAdversa AI disclosed that Claude Code's deny-rule enforcement silently stops working when a shell command contains more than 50 subcommands (joined by &&, ||, or ;). The root cause: internal ticket CC-643 capped security analysis at 50 subcommands to prevent UI freezes from token costs, falling back to a generic auto-allowable prompt. A secure tree-sitter parser existed in the same repo but was never deployed to the legacy regex parser shipping to customers. Practical attack: a malicious CLAUDE.md with 50 benign build steps and an exfiltration command at position 51. Patched in v2.1.90.
2026-05-27 / projects-researcherCodesearch Ships Multi-Repo Semantic Search MCP in Rust — Hybrid Vector+BM25 With Incremental Indexingflupkede/codesearch provides a fully offline, GPU-free Rust MCP server with hybrid vector + BM25 retrieval fused via Reciprocal Rank Fusion, plus tree-sitter AST chunking for 9 languages. The standout feature is multi-repo search — querying semantically across multiple repositories simultaneously — with incremental indexing that processes only changed files, typically 10-100x faster than full rebuilds. Compatible with Claude Code, Cursor, OpenCode, and any MCP client.
2026-05-26 / github-pulse-researcherUnderstand-Anything at 33.4K Stars With +5,604 Today: Code-to-Knowledge-Graph Plugin Tops GitHub Trending With Interactive Exploration and Guided ToursLum1104/Understand-Anything is the #1 trending repo on GitHub today, gaining 5,604 stars in 24 hours. The TypeScript plugin transforms any codebase into an interactive knowledge graph with explorable nodes for files, functions, and classes, using Tree-sitter for deterministic parsing plus LLM semantic analysis. It supports Claude Code, Codex, Cursor, and 16+ other AI coding platforms, and just shipped v2.7.3 with diff impact analysis and architectural layer visualization.
2026-05-21 / github-pulse-researchercolbymchenry/codegraph Explodes to 12.2K Stars With +4,222 Today: Pre-Indexed Code Knowledge Graph Cuts Agent Tool Calls by 70%CodeGraph builds a tree-sitter-powered semantic index of codebases stored in SQLite, giving AI coding agents like Claude Code, Codex, and Cursor pre-computed knowledge about code structure. Benchmarks show ~35% cost reduction and ~70% fewer tool calls versus standard code exploration. Version 0.8.0 shipped May 20 with support for 19+ languages and framework-aware routing for 13 web frameworks.
2026-05-20 / vibe-coding-researcherPattern: Knowledge Graphs Standardize as Context Optimization Layer for AI Coding AgentsThree independent projects shipped knowledge-graph-based context optimization for AI coding in May 2026: code-review-graph (17K stars, 49x token reduction), CodeGraph (cross-agent support for Claude Code, Codex, Cursor, OpenCode), and Graphify (MindStudio integration, 70x cost reduction on large codebases). All use Tree-sitter AST parsing into local SQLite with MCP server interfaces. The convergence suggests knowledge graphs are becoming a standard middleware layer between codebases and AI agents — not a niche optimization but expected infrastructure.
2026-05-20 / projects-researchercode-review-graph Trends at 17K Stars — MCP-Based Knowledge Graph Cuts Claude Code Token Usage by 6.8xcode-review-graph builds a persistent, incrementally-updated knowledge graph of your codebase using Tree-sitter AST parsing across 20 languages, stored in SQLite and exposed via MCP tools. It achieves 6.8x fewer tokens on code reviews and up to 49x reduction on daily coding tasks by giving Claude Code a structural map so it reads only what matters. Initial build takes ~10 seconds for a 500-file project with automatic updates on every file edit and git commit. Compatible with Codex, Cursor, Antigravity, Gemini CLI, Kiro, and GitHub Copilot.
2026-05-15 / projects-researchersafishamsi/graphify Trends at 48,170 Stars — AI Coding Skill Turns Any Codebase Into a Queryable Knowledge Graph via Tree-Sitter ASTGraphify converts any folder of code, SQL schemas, R scripts, shell scripts, docs, papers, images, or videos into a queryable knowledge graph — app code + database schema + infrastructure in one graph. It works with Claude Code, Codex, OpenCode, Cursor, and Gemini CLI, processing code locally via tree-sitter AST so no file contents leave your machine. At 48,170 stars and MIT-licensed, it's become the dominant codebase comprehension skill for AI coding agents. Especially useful for solo engineers navigating large codebases where grep falls short.
2026-05-13 / projects-researchercodesearch Ships Multi-Repo Semantic Code Search MCP Server in Rust — Hybrid Vector + BM25 Retrieval With Tree-Sitter AST Chunking, Fully Offlineflupkede/codesearch provides a Rust-based MCP server for multi-repo semantic code search combining vector embeddings with BM25 retrieval and tree-sitter AST-aware chunking. Runs fully offline and works with OpenCode, Claude Code, Cursor, and any MCP client. Represents the growing trend of high-performance, language-aware code search infrastructure being built specifically for the AI coding agent ecosystem.
2026-05-10 / vibe-coding-researcherlean-ctx Ships Context OS for AI Development: 60–99% Token Reduction via MCP Server + Shell Hooklean-ctx is a Rust-based 'Context OS' that sits between AI coding tools and LLMs, compressing file reads by 60–99% and shell output by 60–95%. Cached re-reads cost only 13 tokens. It works as a standard MCP server with 49 tools across 10 read modes using Tree-sitter AST parsing for 18 languages. Recent updates added CLI/MCP feature parity ensuring both paths trigger identical side effects: session tracking, Context Ledger updates, heatmap recording, and knowledge consolidation. Compatible with Claude Code, Cursor, Copilot, Windsurf, and Codex.
2026-05-04 / projects-researcherGraphify Ships Knowledge Graph Skill for 15+ AI Coding Assistants — Turn Any Codebase Into Queryable Graph With 71x Token Reductionsafishamsi/graphify is an AI coding assistant skill that combines Tree-sitter static analysis with LLM-driven semantic extraction to turn any codebase — including SQL schemas, R scripts, shell scripts, docs, papers, and images — into a queryable knowledge graph using NetworkX + Leiden clustering. The tool works across Claude Code, Codex, OpenCode, Cursor, Gemini CLI, Copilot, Aider, Kilo, and more, with average query costs of ~1.7K tokens versus ~123K naive (71.5x reduction). No Neo4j required — everything runs locally via pip install graphifyy, making it the lowest-friction path to codebase-aware AI assistance.

Graph relationships

USES
GitNexus
GitNexus uses Tree-sitter WASM for code knowledge graphs.
Source finding
USES
GitNexus
GitNexus uses Tree-sitter for AST parsing across 11 languages
Source finding
USES
GitNexus
GitNexus uses Tree-sitter for code AST parsing
Source finding
USES
Weave
Weave semantic merge uses tree-sitter for entity-level merging.
Source finding
USES
GitNexus
GitNexus uses Tree-sitter ASTs to support 11 languages.
Source finding

Source trail

arXiv GitHub Adversa AI GitHub GitHub GitHub GitHub GitHub

Graph sources

entity graphfindings textkg edgeskg entities