AI / LLM Integration
Machine-readable docs for AI coding assistants.
Both `.txt` endpoints regenerate on every docs build. The corpus tracks the published site.
Endpoints
| Endpoint | Purpose |
|---|---|
/llms.txt | Indexed link list of every doc page. Drop into any LLM context. |
/llms-full.txt | Same index plus embedded per-page summaries. Full corpus in a single fetch. |
context7.com/websites/reelkit_dev | Live-indexed Context7 manifest. Plug in via the @context7 MCP server. |
Quick start
Agent prompt
Paste an endpoint URL into the agent to ground answers in current docs.
reelkit.dev/llms-full.txt How do I wire vertical-feed gestures?Context7 MCP server
Install the @context7 MCP server. The agent picks up the ReelKit manifest automatically and fetches docs on demand. Mention reelkit in your prompt.
Direct ingestion
Fetch programmatically:
curl -s https://reelkit.dev/llms.txt
curl -s https://reelkit.dev/llms-full.txtIn-browser agents (WebMCP)
Every page on reelkit.dev registers WebMCP tools, so an AI agent running in your browser reads the docs directly instead of scraping the rendered page.
| Tool | Purpose |
|---|---|
list_pages | Every docs page with its title, URL and group. |
search_docs | The site search by query, with optional locale and framework. Up to 20 results, section anchors included. |
get_page | One page as markdown: the English text llms-full.txt holds for it. |
open_page | Opens a page in the current tab, in the language you are reading, optionally at a section. |
WebMCP is a draft standard. On reelkit.dev it works in Chrome 149 to 156 through an origin trial; anywhere else, Chrome needs chrome://flags/#enable-webmcp-testing. Browsers without it load nothing extra. The tools only read the public docs: no cookies, no storage, no requests beyond this site.
What gets indexed
- Getting started + Installation
- Core engine guide + API
- React, Vue, Angular bindings (guide + API + reel-player + lightbox + stories-player)
- Stories core engine
- SSR notes
- Troubleshooting
- Changelog
Why?
AI assistants lag behind library changes. Generated code references stale APIs. These endpoints update with every doc release, so suggestions match current behavior instead of last quarter's.