Introducing Wunderland: The Free Open-Source AI Agent Framework

What is Wunderland?

Wunderland is a free, open-source TypeScript framework for building and deploying autonomous AI agents. Built as an enhanced fork of OpenClaw and deeply integrated with AgentOS, it provides everything you need to create agents that can browse the web, research topics, post across social media, manage accounts, send messages, and make decisions — all autonomously.

Why Another Agent Framework?

The AI agent ecosystem is exploding, but most frameworks fall into two camps: either too simple (glorified prompt chains) or too complex (enterprise-only, closed-source, expensive). Wunderland sits in the sweet spot:

• Free and open-source under the Apache-2.0 license — use it commercially, fork it, modify it
• TypeScript-first — full type safety, modern tooling, great DX
• Self-hostable — run entirely offline with Ollama, no API keys required
• Production-ready — 5-tier security, HMAC output signing, cost guards, audit logging

Core Capabilities

HEXACO Personality Engine: Every Wunderland agent has a six-factor personality model (Honesty-Humility, Emotionality, eXtraversion, Agreeableness, Conscientiousness, Openness) that drives its mood, posting style, decision-making, and social interactions. Two agents with different HEXACO profiles will respond to the same situation in genuinely different ways. Unlimited RAG Memory: Five-tier memory architecture — working memory for current context, long-term semantic memory for knowledge, episodic timeline for chronological recall, GraphRAG for entity relationships, and shared agency memory for multi-agent collaboration. 51+ Extensions: Browser automation with Playwright, web scraping with proxy rotation and CAPTCHA solving, 28 messaging channels (WhatsApp, Telegram, Discord, Slack, and 24 more), 13 LLM providers, voice synthesis with ElevenLabs, and much more. 5-Tier Security: From "dangerous" (no guardrails, for testing) to "paranoid" (pre-LLM classification, dual-LLM auditing, HMAC output signing, cost guards, and full audit logging).

Getting Started

# Install globally
npm install -g wunderland

# Create a new agent
wunderland init my-agent

# Start chatting
wunderland chat

Or self-host entirely offline:

# Pull a local model
ollama pull llama3.2

# Start with local inference
wunderland start --provider ollama --model llama3.2

Built by Rabbit Hole Inc

Wunderland is developed and maintained by Rabbit Hole Inc, the team behind the Rabbit Hole AI platform. We believe autonomous AI agents should be accessible, safe, and open. The entire framework is licensed under Apache-2.0 and available on GitHub.

Join our Discord to connect with other builders, or check out the documentation to dive deeper.