zzet/gortex

What it does

Gortex parses source repositories with tree-sitter and compiler-grade resolvers, then builds a persistent knowledge graph of functions, classes, call chains, HTTP/gRPC/GraphQL routes, message topics, and cross-service contracts. AI coding agents and IDEs query that graph through MCP, the CLI, or an HTTP /v1 API instead of reading whole files, which the project reports can cut token usage by up to 50x. It supports 257 languages and multi-repository graphs by default, and runs fully locally as a single static binary with no external dependency chain. See the feature catalog and language support.

Quick start

Install the binary (macOS/Linux):

curl -fsSL https://get.gortex.dev | sh

The Windows alternative uses irm https://get.gortex.dev/install.ps1 | iex. The installer detects OS/arch and verifies SHA256 and a cosign signature; other install methods are documented in the installation guide.

Then set up and index repositories:

gortex install                       # one-time machine setup
gortex daemon start --detach         # background daemon
gortex track ~/projects/myapp        # add a repo
cd ~/projects/myapp && gortex init   # configure the repo for agents

The result is that detected AI assistants use graph queries for the tracked repository. A longer walkthrough is in the onboarding guide.

Operating constraints

• Building from source requires Go 1.26+ and CGO for the tree-sitter C bindings (README).
• The daemon keeps a long-lived process serving all IDE windows through a shared SQLite graph store.
• Semantic search is on by default with an embedded GloVe-50d model; MiniLM, Ollama, or OpenAI embeddings are opt-in.
• Telemetry is off by default and only aggregates anonymous tool counts when enabled; DO_NOT_TRACK is honored.
• The README explicitly notes that Gortex has no cryptocurrency token or coin.
• Licensed under Apache-2.0.

HotRepo's take

Gortex entered the Go daily list on 2026-11-05 with 33 new stars and a best position of 16 that day, at 1,995 total stars and 194 forks. Best GitHub daily rank and historical star-curve samples are not available in the record, so a single daily snapshot supports only that day's observation rather than a broader growth trend. The project's own scale benchmarks report indexing the Linux tree (70,333 files) in roughly 3 minutes and VS Code in about 1 minute on an Apple Silicon laptop, though those are author-measured figures rather than independent results.