Skip to content

Knowledge Base 🦞

Upload documents to give your agent specialized knowledge that enhances its conversations and makes it an expert in specific domains. Knowledge integrates with your OpenClaw agent's existing intelligence and Hi's character layer.

Supported Formats

  • PDF — research papers, books, manuals
  • Text / Markdown — articles, guides, documentation
  • JSON / CSV — structured data and databases

How It Works

Upload your documents and Hi processes them into a searchable knowledge base. When your agent receives a message on any OpenClaw channel, it retrieves relevant knowledge to inform its response — giving it accurate, domain-specific answers instead of generic output.

Upload flow

  1. Upload documents through the Hi dashboard
  2. Content is extracted, chunked, and embedded as vectors
  3. Knowledge becomes available to your agent across all OpenClaw channels

Memory System

Beyond uploaded knowledge, your agent remembers conversations across all OpenClaw channels:

  • Persistent memory — your agent recalls past interactions with each user
  • Cross-channel sync — a conversation started on iMessage is remembered on WhatsApp, Discord, Telegram, Slack, and everywhere else your OpenClaw agent lives
  • Preference learning — your agent adapts to individual users over time
  • Relationship building — interactions deepen as your agent remembers context, preferences, and history

Knowledge + Character Layer

Knowledge integrates with Hi's character layer. Your agent doesn't just retrieve facts — it communicates them in its own voice, personality, and style as defined in its OpenClaw identity files (SOUL.md, IDENTITY.md). A knowledgeable agent with Hi installed delivers expertise as a recognizable character, not a generic chatbot.