Skip to main content

Knowledge for your in-product agent

In-product agents are AI assistants embedded directly in your application, typically as a sidebar, panel, or full-page chat. They let people operate your product through AI instead of hunting through menus and manuals: beyond answering questions, they can execute tools, take actions, and reason through multi-step tasks on behalf of your users, which removes friction and helps users get more out of your product. Common patterns:

  • Product help: answer "How do I...?" questions using your documentation.
  • Data and operations: run queries, inspect entities, or pull recent activity using your APIs.
  • Guided workflows: walk users through configuration, setup, or debugging steps.
  • Actions with approval: create resources, update settings, or trigger jobs with user confirmation.

Whatever the pattern, these agents need access to your knowledge. Without it, they cannot answer product questions, make sense of what their tools return, or know what your product can do in the first place. Knowledge base search in AI agents shows how much of an agent's usefulness depends on it.

Ways to build this with Kapa​

  • Build your agent however you like, and give it access to Kapa's retrieval. Kapa indexes all your knowledge sources into one knowledge base and lets your agent search it whenever it needs context. You own the agent, built with whatever framework you prefer, and Kapa is one of its tools, added via the hosted MCP server or the Retrieval endpoint.
  • Or build on the Agent SDK if you do not want to own the agentic loop yourself: the simplest and fastest way to get an in-product agent. It is an end-to-end package with orchestration, streaming chat UI (React components or headless hooks), custom tools with user approval, theming, and knowledge base retrieval built in, frontend-first, so there is no agent backend to deploy. Install the Agent SDK integration skill into your AI coding agent and it scaffolds the whole integration interactively.

What to index​

Everything the agent should be able to answer about your product: your documentation via Website Crawl, API specifications via OpenAPI, and code, READMEs, and issues via the GitHub connectors. Start with Index your first source.

Resources​