Maximize Your AI Agent's Context Window

An MCP server that enables AI agents to read, navigate, and edit codebases through compact structural context while preserving source-level names.

Get StartedRead DocumentationTry Live Concept Demo

From Source Tree to Compact MCP Results

Below is a progressive visual sequence demonstrating how Project Graph compiles structure contexts.

  1. PHASE 1

    Source Tree

    Identify code files in your workspace, imports, and dependencies.

  2. PHASE 2

    AST Parser

    Parse files to produce structural facts like classes and functions.

  3. PHASE 3

    Compact Skeleton

    Create an AI-efficient structure substituting long names with short symbols.

  4. PHASE 4

    Focus Zone

    Isolate code changes in your git diff or target focus area.

  5. PHASE 5

    MCP Result

    Deliver optimized structured context straight to the AI agent.

SOURCE TREE·source files and changesets loaded

Identify code files in your workspace, including imports and dependency relations.

Project Graph MCP Timeline VisualizationAn interactive diagram showing the phases: Source Tree, AST Parser, Compact Skeleton, Focus Zone, and MCP Result.WorkspaceAST TreePWMUSkeleton GraphPWMUFocus Zonegit diffMCP Result
↔ Scroll horizontally to view full diagram
0.00s / 10.00s

Core Capabilities

AST Skeletons

Compact structural maps provide file profiles and legends for exported symbols.

Compact Mode

Keep source-level names while delivering compact structural context and reversible edits.

Multi-Language Support

Parse JavaScript ASTs through Acorn and use custom adapters for TypeScript, Python, and Go.

Custom Rules Analysis

Run project-specific style and pattern checks selected from package dependencies.