Kivgraph: Local Semantic Code Intelligence for Multi-Repository Projects

Overview

Kivgraph is a local Model Context Protocol server designed to provide deep semantic code intelligence across multiple repositories. It specifically supports Go, TypeScript, and Rust projects. Unlike traditional search tools that rely on simple text matching, Kivgraph indexes code by resolving edges using language-specific checkers such as go/types, the TypeScript checker, and rust-analyzer to create an immutable graph. This approach ensures precise answers regarding code relationships, enabling users to confidently identify references, understand transitive impacts, find cross-repository consumers, and even prove the absence of specific code interactions.

Benefits

Kivgraph offers several distinct capabilities designed to enhance code analysis. It enables the discovery of code consumers that span across multiple repositories, providing a holistic view of project interdependencies. By leveraging language-specific type checkers, Kivgraph performs accurate code relationship analysis that goes beyond keyword matching. Users can access a read-only 3D graph visualization of the indexed code through a dedicated web viewer. The tool allows for the tracing of dependencies and the analysis of the blast radius of potential code changes, helping developers assess risk before implementation. It provides cross-repository semantic code intelligence specifically tailored for Go, TypeScript, and Rust.

Use Cases

Kivgraph is particularly useful in several scenarios. It helps developers understand intricate code dependencies and call graphs across multiple projects. It allows teams to evaluate the impact of proposed code changes prior to implementation to minimize disruption. The tool also helps in identifying all usages of a specific symbol, including those located in other repositories, with high precision.

Additional Information

The core differentiator of Kivgraph is its reliance on immutable graphs constructed via language-specific checkers. This method ensures that the semantic relationships identified are accurate and verifiable, offering insights that simple text searches cannot provide. By operating locally and utilizing the Model Context Protocol, it integrates seamlessly into developer workflows while maintaining data privacy and control.