gortex

Gortex is a code-intelligence engine aimed at AI coding agents and IDEs. It parses source code into a persistent, provenance-tiered knowledge graph of functions, classes, call chains, HTTP routes and cross-service contracts, then exposes that graph through a CLI, an MCP server, an HTTP API and a web UI. Key capabilities described in the README: - Language coverage: parsing of 257 languages/grammars via tree-sitter AST analysis, with compiler-grade resolution for a subset including Python, TypeScript/JavaScript, PHP, C#, Go, C, C++, Java, Kotlin, Swift, Zig, Rust, Ruby, Elixir, OCaml and Haskell. Jupyter and Databricks notebooks are also handled. - Multi-repository support: multiple repos can be indexed into one graph, with cross-repo edges, contract matching and impact analysis. - Agent integration: the project states out-of-the-box support for 19-20 coding agents (Claude Code, Cursor, Windsurf, VS Code/Copilot, Continue.dev, Cline, OpenCode, Codex CLI, Gemini CLI, Zed, Aider and others), configured by a single `gortex init` command. - MCP surface: a configurable set of MCP tools (README mentions 175 configurable tools, 16 resources and 3 prompts) covering symbol lookup, call chains, blast radius, dataflow, clone detection and refactoring. - Token efficiency: the README claims up to 50x fewer tokens per response versus naive file reads, plus an additional ~27% reduction from its GCX1 wire format. A `gortex savings` command reports estimated tokens and cost avoided. - Search: semantic search is enabled by default using an embedded GloVe-50d model with FTS5/BM25 and vector fusion; optional MiniLM, Ollama or OpenAI backends. - Local-first design: a single static binary for macOS, Linux and Windows with no external dependency chain; a long-lived daemon serves multiple IDE windows and watches the filesystem. - Additional features: speculative edit previews, live editor overlays for unsaved buffers, PR triage and review tooling, an HTTP `/v1/*` API with SSE events, and optional LLM provider integrations (local llama.cpp, Anthropic, OpenAI, Ollama, Gemini, Bedrock, DeepSeek and CLI subprocesses). - Privacy: telemetry is off by default and honours DO_NOT_TRACK. Installation is via a shell or PowerShell one-liner, with Homebrew, deb/rpm/apk, scoop and source builds also documented. Building from source requires Go 1.26+ and CGO. The project is licensed under Apache 2.0. The README includes benchmark tables for large repositories such as the Linux kernel and VS Code, but these are self-reported figures from the project's own documentation.