claudecode-rlm

Unlimited context for Claude Code - Knowledge graph-based context storage with surgical precision code analysis.

Author: Michael Thornton (tekcin@yahoo.com) Repository: https://github.com/tekcin/claudecode-rlm

Features

• TRUE Unlimited Context - Proxy server auto-stores ALL conversations, auto-injects relevant context
• Surgical Precision - Deep code analysis for function, class, and symbol-level understanding
• Live File Tracking - Real-time monitoring of code changes with automatic re-analysis
• Repository Loading - Bulk load entire codebases into the knowledge graph
• Web Console - Full management GUI with file browser, symbol explorer, and graph visualizer
• Docker Ready - Production-ready containerization with Nginx reverse proxy
• Knowledge Graph - Hierarchical storage (Document → Section → Chunk → Entity → Symbol)
• 74x Faster Reads - LRU cache with inverted index
• Entity Extraction - Automatic extraction of code elements, files, concepts
• Multi-Language Support - TypeScript, JavaScript, Python, Go, Rust

Quick Start (30 seconds)

# 1. Install and setup
git clone https://github.com/tekcin/claudecode-rlm.git
cd claudecode-rlm
npm install && npm run build && npm run setup

# 2. Start proxy (keep running)
npm run proxy

# 3. Open NEW terminal and use Claude Code
claude

That's it! Works with Claude Max subscription (no API key needed) or API key.

Web Console

Open http://localhost:3456/web to browse files, symbols, and the knowledge graph.

Alternative: Docker

git clone https://github.com/tekcin/claudecode-rlm.git
cd claudecode-rlm
docker-compose up -d

Alternative: Global Install

npm install -g claudecode-rlm
claudecode-rlm-proxy

Proxy Mode (Recommended)

The proxy server intercepts ALL API calls for complete automatic context:

1. Start the proxy

claudecode-rlm-proxy

2. Configure Claude Code

export ANTHROPIC_BASE_URL=http://localhost:3456

What happens automatically:

• Every user message → Stored in knowledge graph
• Every assistant response → Stored in knowledge graph
• Before each request → Relevant past context injected
• Streaming supported → Works with Claude Code's streaming
• Live tracking → Auto-watches your working directory

Web Console

Access the full management console at http://localhost:3456/web

Features

Building the Web Console

cd web
npm install
npm run build

Repository Loading

Load entire codebases into the knowledge graph:

Via Web Console

1. Open http://localhost:3456/web
2. Go to Upload Manager
3. Enter repository path and file patterns
4. Click Load Repository

Via API

curl -X POST http://localhost:3456/repo/load \
  -H "Content-Type: application/json" \
  -d '{
    "rootDir": "/path/to/repo",
    "patterns": ["**/*.ts", "**/*.js", "**/*.py"]
  }'

Via CLI Menu

claudecode-rlm-menu

Surgical Precision Code Analysis

Get precise code context for surgical edits:

Find Symbol

curl "http://localhost:3456/code/symbol?name=processRequest"

Get Context at Location

curl "http://localhost:3456/code/context?file=src/index.ts&line=42"

Find All Usages

curl "http://localhost:3456/code/usages?name=GraphStorage"

Get Call Hierarchy

curl "http://localhost:3456/code/callers?name=ingestFile&direction=both"

Live File Tracking

Real-time monitoring of code changes:

Start Watching

curl -X POST http://localhost:3456/live/start \
  -H "Content-Type: application/json" \
  -d '{"rootDir": "/path/to/watch"}'

Get Live File Content

curl "http://localhost:3456/live/content?path=src/index.ts"

Subscribe to Changes (SSE)

curl http://localhost:3456/live/updates

Docker Deployment

Basic Deployment

docker-compose up -d

Production with Nginx + SSL

docker-compose -f docker-compose.prod.yml up -d

Development with Hot Reload

docker-compose -f docker-compose.dev.yml up

See DOCKER.md for full deployment guide.

API Reference

Proxy Endpoints

Repository Endpoints

Code Context Endpoints

Live File Endpoints

Upload Endpoints

Graph API Endpoints

MCP Mode

For manual control via MCP tools:

Add to Claude Code config

{
  "mcpServers": {
    "claudecode-rlm": {
      "command": "claudecode-rlm"
    }
  }
}

Available Tools

Architecture

Knowledge Graph Structure

Document (conversation turn)
├── Section (topic/header group)
│   ├── Chunk (~300 chars, searchable)
│   │   ├── Entity (function name)
│   │   ├── Entity (file path)
│   │   └── Entity (concept)
│   └── Chunk
│       └── [FOLLOWS] → next chunk
├── File (source code file)
│   ├── Symbol (function/class/method)
│   │   ├── [CALLS] → Symbol
│   │   ├── [REFERENCES] → Symbol
│   │   └── [CONTAINS] → Symbol
│   └── Import
│       └── [IMPORTS] → File
└── Section

Node Types

Relationship Types

Storage Location

.claudecode-rlm/graph/{sessionID}/
├── nodes/*.json    - Content nodes
├── edges/*.json    - Relationships
└── indexes/*.json  - Search indexes

Configuration

Environment Variables

Performance

Development

# Install dependencies
npm install

# Build TypeScript
npm run build

# Build web console
cd web && npm install && npm run build && cd ..

# Start MCP server
npm start

# Start proxy server
npm run proxy

# Start with Docker
npm run docker:up

License

MIT

Author

Michael Thornton

• Email: tekcin@yahoo.com
• GitHub: @tekcin

Links

• npm package
• GitHub
• Issues
• Docker Guide