How to use the KX documentation
Learn where to find product documentation, how to confirm the version you need, and how to get quick answers with Ask AI or AI tools.
KX publishes documentation for all KX products and tools on code.kx.com. Each product has its own site on code.kx.com with its own navigation. This page shows how to find what you need, how to ask questions and search with Ask AI, and how to give AI tools and agents access to the docs.
Find the product you need
Start on the Product documentation home page. It links to every product site.
- Use the navigation tabs at the top of a product site to move between its main sections.
- Use the On this page list to jump to a section of a long page.
- Use the search box on a product site to search that product.
- Use the search on home to do a site-wide Google search.
Check the version
Some products publish more than one version. When a page URL contains a version number, such as /pykx/4.1/, the page describes that version. A link without a version, such as /pykx/, goes to the current release.
Check the version before you copy code, settings or commands, especially if you use an older release.
Ask AI
Ask AI is an experimental assistant for technical questions about KX products. Select Ask AI to open it in a side panel.
- Answers come from KX documentation (latest versions only) and technical sources such as Tutorials, Academy, Forum, and KX Github repositories.
- Answers can be wrong. Check the source pages an answer refers to before you rely on it.
- Start with an example question, or ask a specific, complete question. For example, "How do I install KDB-X Python?" works better than "install help".
- You can also post errors and logs and Ask AI to diagnose and troubleshoot the issue, or ask for help creating configuration files.
- Rate an answer with the feedback buttons. Your ratings help KX improve the answers.
- Use search mode to find source pages for your question, rather than a written answer.
Tip
Search mode searches all the sources in Kapa, so you can use it as a global search for KX developer resources.
Ask AI is not a support channel. For product issues, see Product support.
Warning
Do not include credentials, personal data, or confidential information in your questions.
Use the docs with AI tools and agents
KX publishes the documentation in forms that AI tools can read directly.
Connect an MCP client to Ask AI
The KX Docs MCP server is available at:
https://kx.mcp.kapa.ai
Connect it from an MCP client, such as an AI desktop app or an IDE (integrated development environment) extension. The server uses OAuth, a standard sign-in method, so your client opens a browser window to sign in the first time it connects. You need a Google or GitHub account to sign in. KX uses the account only to enforce per-user rate limits and prevent abuse.
If your client cannot sign in, use the retrieval endpoint or the server below instead.
Note
A second docs MCP server, kx-docs-mcp, authenticates with a bearer token instead of OAuth. The easiest way to add this server to your setup is to install the KX Plugins for Claude Code and Codex. For more information, see the kx-skills repository.
Both MCP servers provide these tools:
| Tool | Purpose |
|---|---|
search_kx_knowledge_sources |
Semantic search over KX documentation. Takes query, a complete sentence. Returns ranked passages, each with its source_url and Markdown content. |
get_kx_knowledge_documents |
Fetches whole pages by their exact source_url. Takes urls (1 to 100), plus optional max_chars_per_document (default 50000), page_size (1 to 5), and page. Use it to read a full page after a search. |
give_feedback |
Sends feedback to the KX team about the documentation, a product, or the server. Takes message and category (documentation, product, mcp_server, or other), plus optional context, severity, and tool_name. |
A typical flow is to search first, then fetch the pages you need in full.
Set up an MCP client
Add an MCP server to your client with the URL https://kx.mcp.kapa.ai. The client then opens a browser window so you can sign in.
Tip
You can also use the Use MCP button in the Ask AI sidebar to connect your preferred client.
Connect AI tools to running KX products
Some KX products have their own MCP servers. These connect an AI client to a running product, so the client can query or work with that product's data. They do not search the documentation.
Find pages with llms.txt
llms.txt is an index of KX documentation that lists one line per page, with a short description.
Give your agent this file first, then follow the links to the pages it needs.
Search with the retrieval endpoint
The retrieval endpoint searches KX official documentation, community forum, and blog content. It needs no authentication.
curl -s "https://kx-docs-retrieval.kapa-kx.workers.dev/?q=How%20do%20I%20install%20KDB-X%3F"
The response is a JSON array of objects. Each object has a content field with a passage and a source_url field with the page it came from. Results are ordered by relevance.
The endpoint returns passages, not whole pages. Open the source_url page when you need the full context. The query goes to an external service, so do not include confidential information in it.
Get help
- Product support: Support options for each product, including the KX Support Portal for commercial licenses.
- KX Academy: Courses and forums for community support.
- KX Forum: Community forum.
- KX Slack community: Community discussion.
- KX Developer Center: Developer resources.