Code Graph FAQs

Question 1

Which programming languages does Code Graph support?

Accepted Answer

Code Graph natively supports TypeScript (using the official compiler API) and Python (using the standard library `ast` module). It can ingest both languages directly from source files without requiring a separate build step.

Question 2

How does Code Graph benefit AI agent development?

Accepted Answer

It provides an MCP server with specific tools for AI agents to perform fuzzy symbol lookups (`graph.resolve_symbol`), find inbound references (`graph.references`), discover related code (`graph.related`), and calculate the impact of code changes from a patch file (`graph.impact_from_diff`). It also allows streaming code snippets directly into an agent's context window.

Question 3

What is Code Graph?

Accepted Answer

Code Graph is a lightweight tool that builds a code graph from TypeScript and Python source code. It then exposes this graph via an MCP (Multi-Client Protocol) server, enabling agentic clients like Claude Desktop to query and interact with your codebase efficiently.

Question 4

Does Code Graph require a build step for code ingestion?

Accepted Answer

No, Code Graph is designed for lightweight, no-build ingestion. It parses TypeScript directly via the official compiler API and Python via the `ast` module, making it quick and easy to get started without complex setup or heavy indexers.

Question 5

What kind of code insights and features does Code Graph offer?

Accepted Answer

Code Graph extracts symbols (functions, classes, methods, variables, modules) and edges (import, call, member_of, defines). Its MCP server offers fuzzy symbol lookup, tracing inbound references, identifying k-neighbors, calculating the impact of code changes from patch files, and providing `code://file` resources to stream specific code snippets.

Code Graph

Code Graph

主要功能

使用案例

主要功能

使用案例