Kapa.ai Documentation Assistant
The Zephyr documentation website is augmented with an AI assistant powered by Kapa.ai. The assistant is trained on Zephyr’s own documentation, source code, and GitHub activity (see Data sources below) and can answer free-form questions in natural language, citing the documentation pages and resources it used to build each answer.
The same knowledge base is also exposed through a Model Context Protocol (MCP) server, so that external AI assistants and IDEs can query Zephyr knowledge directly (see Using the MCP server).
Note
Answers are AI-generated and may contain inaccuracies. Always verify important information against the official documentation, and treat the assistant as a starting point rather than an authoritative source.
Using the chatbot
The chatbot is available from every page of the documentation:
Click the chatbot button displayed at the bottom of the page, or
Press Ctrl+K (Cmd+K on macOS) to open it from anywhere.
The assistant opens in a sidebar where you can ask questions in plain language, for example “How do I configure a UART device in the devicetree?” or “What is the difference between a workqueue and a thread?”. Answers include links back to the relevant documentation pages, GitHub issues, or source files so you can dig deeper.
The first time you use the assistant, a one-time consent screen explains how the feature works and what data is collected. See the Privacy section below.
Tip
You don’t have to ask your questions in English. The assistant detects the language you use and responds in the same language, so you can interact with it in your preferred language (English, Chinese, Spanish, Japanese, French, German, and many others).
AI-powered search
In addition to the chatbot, Kapa.ai can power the documentation search box:
Click the gear (Search settings) icon next to the search field.
Select Kapa AI search from the menu.
Type your query and press Enter.
Instead of the built-in keyword search, your query is answered by the AI assistant. Your selection is remembered, so subsequent searches keep using the engine you picked until you change it back.
Using the MCP server
Kapa.ai also exposes the Zephyr knowledge base as an MCP server. MCP is an open standard that lets AI assistants and AI-enabled IDEs connect to external tools and knowledge sources. By adding the Zephyr MCP server to your assistant, you give it the ability to query up-to-date Zephyr documentation, source code, and GitHub activity on your behalf.
The server is available at the following endpoint, using the streamable HTTP transport:
https://zephyrproject.mcp.kapa.ai
The exact configuration steps depend on the MCP client you use. Most clients read a JSON
configuration file with an mcpServers section similar to the following:
{
"mcpServers": {
"zephyr-docs": {
"url": "https://zephyrproject.mcp.kapa.ai"
}
}
}
Some clients instead expect the server to be launched through a local stdio bridge such as mcp-remote:
{
"mcpServers": {
"zephyr-docs": {
"command": "npx", "args": ["-y", "mcp-remote", "https://zephyrproject.mcp.kapa.ai"]
}
}
}
Refer to the documentation of your specific MCP client (AI coding assistant, IDE extension, etc.) for the precise location and syntax of its configuration.
Tip
Connecting the MCP server to an AI coding assistant lets it answer Zephyr questions and ground its code suggestions in the project’s actual documentation and source, rather than relying solely on its training data.
Data sources
Kapa.ai builds its answers exclusively from the following Zephyr project sources, which are synchronized at the indicated frequencies:
Source |
Details |
Refresh frequency |
|---|---|---|
Source code |
|
Hourly |
API reference |
Daily |
|
Devicetree bindings |
Daily |
|
Main documentation |
https://docs.zephyrproject.org/latest/ (excluding the API reference and Devicetree bindings sections) |
Daily |
Project wiki |
Daily |
|
GitHub issues |
|
Every 5 minutes |
GitHub pull requests |
|
Every 10 minutes |
For the authoritative and most up-to-date configuration, refer to the Kapa.ai page on the Zephyr project infrastructure wiki.
Privacy
The assistant uses your questions to provide answers based on the Zephyr documentation, source code, and GitHub issues and pull requests. Questions and interactions may be anonymously collected and analyzed to help improve the documentation by identifying areas that need clarification. No personally identifiable information is collected.