Turn Claude Code Into
A Research Team

Not just a chatbot. Multi-agent orchestration with 32+ MCP tools and 9 LLM providers.

32+ MCP Tools
9 LLM Providers
6 MCP Servers
2 Orchestration Patterns
Get Started View on GitHub

Why Dreamwalker?

🎯

Multi-Agent Orchestration

Deploy Beltalowda's 3-tier synthesis (8 agents → mid-tier → executive) or Swarm's specialized parallel agents for domain-specific research.

🔌

9 LLM Providers

Unified interface across Anthropic, OpenAI, xAI, Mistral, Cohere, Gemini, Perplexity, Groq, and HuggingFace. Switch providers without code changes.

🛠️

32+ MCP Tools

Chat completion, vision analysis, image generation, web search, data fetching, caching, citations, TTS, and more—all through MCP protocol.

⚡

Real-Time Streaming

SSE streaming for live progress updates. Watch agents work in parallel, synthesize findings, and generate comprehensive reports.

📊

Data Fetching Suite

Wikipedia, YouTube, Finance, Weather, GitHub, NASA, News, OpenLibrary—10+ data sources with automatic caching and rate limiting.

🔐

Production-Ready

Redis caching, OpenTelemetry observability, rate limiting, retry logic, comprehensive error handling, and full type safety.

6 MCP Servers, 32+ Tools

Each server exposes specialized tools through Claude Code's MCP protocol:

🎯 Unified Orchestrator 3 tools

Beltalowda hierarchical research (8→1→1) and Swarm parallel multi-agent orchestration. Includes task status tracking and cancellation.

🔌 LLM Providers 4 tools

Unified interface for 9 LLM providers. Chat completion, image generation (DALL-E), vision analysis, and model listing.

📊 Data Fetching 10 tools

Wikipedia, YouTube transcripts, Finance data, Weather, GitHub repos, NASA imagery, News articles, OpenLibrary books, and more.

💾 Cache Manager 5 tools

Redis-based caching with TTL, namespaces, key management, statistics tracking, and state persistence.

🔬 Utilities 4 tools

ArXiv paper search, TTS/audio generation, academic citations (APA/MLA/Chicago), and Semantic Scholar integration.

🔍 Web Search 1 tool

Multi-engine web search with SerpAPI, Tavily, and Brave. Auto-selects best available provider with fallback support.

Installation

Get started in under 5 minutes:

1

Clone Repository

Clone the plugin from GitHub:

git clone https://github.com/lsteuber/dreamwalker-mcp.git cd dreamwalker-mcp
2

Install Package

Install in development mode:

pip install -e .
3

Configure API Keys

Copy environment template and add your keys:

cp .env.example .env nano .env # Add API keys
4

Enable in Claude Code

Add to Claude Code settings.json:

{ "mcpServers": { "dreamwalker": { "command": "python", "args": ["-m", "dreamwalker_mcp.mcp.stdio_servers.unified_stdio" ] } } }

Full documentation: README.md on GitHub

Performance Benchmarks

Dreamwalker orchestrators deliver measurable improvements over single-agent workflows:

3.2x Faster Research

Beltalowda parallel agents vs sequential

87% More Comprehensive

Coverage vs single-agent baseline

4.1/5 Quality Score

Expert evaluation (n=50 reports)

<200ms Tool Response

P95 latency with Redis caching