elephant.ai
Proactive AI Assistant for Every Surface
One runtime, every surface, always ready.
Why elephant.ai?
Channel-Native
First-class participant in Lark groups, WeChat chats, CLI, and web dashboard. No app switching needed.
Persistent Memory
Remembers conversations, decisions, and context across sessions. The assistant retrieves relevant history automatically.
Autonomous Execution
ReAct loop runs Think-Act-Observe to completion: web search, code execution, file ops, browser automation.
Built-in Skills
15+ markdown-driven workflows: deep research, meeting notes, email drafting, slide decks, video production, code review.
Multi-Provider LLM
Seamless integration with OpenAI, Claude, DeepSeek, ARK, Gemini, Ollama, and more. No vendor lock-in.
Approval Gates
Proactivity with safety. Risky tool calls require explicit human approval via CLI, web, or chat.
Architecture
Clean port/adapter boundaries. Multi-provider. Observable.
Delivery Layer
CLI, Server (HTTP/SSE), Web Dashboard, Lark Bot, WeChat Bot
Agent Application
Task preparation, coordination, cost accounting, context assembly
Domain (ReAct Loop)
Typed events, approval gates, Think-Act-Observe cycle
Infrastructure Adapters
LLM providers, tools, memory store, RAG, observability