Three core capabilities
PR reviews
Comment
@bugviper review on any pull request to trigger an inline AI review. The agent explores the knowledge graph for context — call chains, class hierarchies, cyclomatic complexity — before flagging bugs and security issues on the exact diff lines.Code search
Search your entire indexed codebase by symbol name, file content, or natural language intent. Full-text search is backed by Apache Lucene inside Neo4j; semantic search uses cosine similarity over 1536-dimension embeddings.
Ask Agent
Ask the AI chat agent natural language questions about your codebase. It reasons across 19 Neo4j tools, cites the source files it used, and shows relevant code inline — without you needing to know Cypher.
The knowledge graph model
When you index a repository, BugViper runs Tree-sitter parsers across every source file and writes the result into Neo4j as a property graph. Each entity in your code becomes a node, and the connections between them become typed relationships:| Node type | What it represents |
|---|---|
Repository | The top-level repository being indexed |
File | A source file (path, language, raw content) |
Function | A function or method, including its source and cyclomatic complexity |
Class | A class definition with its member functions |
Variable | Module-level or class-level variable declarations |
Module | An imported package or module |
CONTAINS, DEFINES, CALLS, IMPORTS, INHERITS — let the review agent trace call chains, check blast radius, and understand inheritance before it writes a single comment.
BugViper calculates cyclomatic complexity for every function at parse time, so the agent can prioritize high-risk areas automatically.
Supported languages
BugViper parses 17 programming languages via Tree-sitter, including Python, TypeScript, JavaScript, Go, Rust, Java, C++, and more. See the full language list.Ready to get started?
The quickstart walks you through indexing your first repository and triggering an AI pull request review in four steps.