gortex

Why we included this project

Teams running AI coding agents against large or multi-repository codebases often find the agent burning through its context window reading files one by one. Gortex tackles that by indexing repositories into an in-memory knowledge graph and handing your agent the relevant slices over MCP, HTTP, or a web UI, so one graph query replaces a stack of file reads. It parses 257 languages with tree-sitter and layers on compiler-grade resolution for a long list of mainstream languages, which makes the returned symbols, callers, and blast-radius answers reliable enough to act on. Beyond token savings, it adds extras like dataflow and taint tracing, clone detection, speculative edit simulation, and graph-grounded PR review. Everything ships as a single static binary that runs fully local, so the whole pipeline stays on your own machines. If your team is fighting token spend or slow agent loops on big codebases, this is worth a close look.