CORE
OPERATIONS.
Arwen builds a living intelligence layer on top of your codebase. It maps relational dependencies and logic flows into an autonomous graph.
MAPPING
Instant synchronization of codebase architecture.
ANALYSIS
Multi-agent validation of logic and security.
SEARCH
Natural language querying of abstract concepts.
SETUP
Rapid deployment across standard IDE platforms.
Installation_Procedures
npm install -g @google/gemini-cliInitialize_Demo
Validate the intelligence layer in real-time.
Launch AnalysisFunction_Search
Smart Filtering & Semantic Query
Search across all nodes, functions, and dependency edges using natural language or exact identifiers. Fuzzy matching and semantic ranking surface the right component instantly.
auth service tokenAST_PARSERdatabase writeExport_&_Download
High-fidelity output for docs, analysis, and sharing
Export the active graph in multiple formats. Filtered JSON preserves the full relational schema. SVG is lossless and scales to any resolution for presentations and documentation.
/arwen export --format json --mode structural --output ./graph.json/arwen export --format svg --filter layer:authDependency_Path_Finder
Shortest path between any two components
Select any two nodes in the graph and instantly trace the shortest dependency path between them. Understand exactly how your system connects — without reading every file.
Guided_Tours
AI-generated walkthroughs for any codebase
Arwen generates step-by-step architectural walkthroughs tailored to your role. New hires get onboarding flows. Engineers get implementation depth. PMs get business context.
JWT signing uses RS256. Session TTL is 24h. Refresh token rotation on every request.
User checkout triggers 3 services: Cart, Payment Gateway, Fulfillment — in sequential order.
This repo contains the core API. The frontend lives in /client. Start by reading AGENT_CORE.
/arwen tour --persona engineer --depth 10Supported_Stack
26+ file types in one unified graph
Beyond code analysis — Dockerfiles, Terraform, SQL schemas, Markdown, and infrastructure configs are all first-class graph nodes alongside your source.
System_Requirements
Claude_Code_Plugin
/plugin marketplace add spooky-may/arwen-intelligence/plugin install arwen-intelligence/arwenCursor_Extension
- 01Search for '@arwen' in the Cursor extension registry.
- 02Install and reload the editor.
- 03Open command palette → 'Arwen: Analyze_Project'.
Copilot_Integration
Requires the Arwen-Copilot bridge extension. Once installed, initialize via the GitHub Copilot chat layer:
@arwen /index_coreGemini_CLI
npm install -g @google/gemini-cligemini loginKnowledge_Graphs
Three layered graph modes — Structural (file/function dependencies), Domain (business logic clusters), Knowledge (semantic concept mapping). All persist locally and sync incrementally.
Agent_Layer
The recursive orchestrator. Manages multi-agent delegation loops and context injection across Planning, Building, Review, and Security agents — each scoped to a graph subgraph.
Task_Orchestration
Tasks are derived from graph state. When a node changes, dependent tasks are invalidated and re-queued. The orchestration layer ensures deterministic execution order using topological sort on the dependency graph.
Security_Models
Local-first. No source code leaves your machine. The graph is built and stored on-device. Cloud sync is optional and encrypted. Static analysis runs in a sandboxed subprocess.