LLM-friendly Documentation
Reference for Y2 documentation artifacts used by coding agents, retrieval systems, and MCP clients.
Y2 publishes build-generated text and MDX artifacts so an AI integration can discover, retrieve, and cite the same documentation shown in Fumadocs.
Choose the smallest useful artifact
Start with /llms.txt. Fetch one page when the task is narrow, use /llms-full.txt only when
broad product context is necessary, and use /api/openapi.yaml for request and response schemas.
How the artifacts are produced
The Vite documentation plugin reads content/docs during a production build. It emits the index,
full corpus, page-level MDX files, and the Fumadocs search index. During local development, the
same plugin generates these responses on request. The OpenAPI reference is generated separately
from public/api/openapi.yaml.
Artifact reference
| Artifact | Content | Use it for |
|---|---|---|
/llms.txt | Plain-text overview and grouped links to every documentation page | Discovery and routing |
/llms-full.txt | Concatenated page titles, canonical paths, and source MDX bodies | Broad grounding or corpus ingestion |
/docs.mdx | Source MDX body for the documentation home page | Targeted retrieval |
/docs/{slug}.mdx | Source MDX body for one documentation page | Task-specific context |
/api/openapi.yaml | OpenAPI contract for public HTTP endpoints | Types, parameters, schemas, scopes, and responses |
These are public documentation artifacts and do not require a Y2 API key.
Source MDX is not rendered prose
Page-level files and /llms-full.txt contain the MDX body without frontmatter. Fumadocs tags such
as Callout, Cards, and APIPage can remain in the text. Generated endpoint pages may point to
public/api/openapi.yaml instead of expanding every schema, so API clients must also read the
published OpenAPI contract.
Retrieve documentation
curl -fsS https://y2.dev/llms.txtUse the grouped links and descriptions to decide which page the task needs.
The deployed files use revalidation caching. Re-fetch them when starting a new integration task or after a Y2 release instead of treating a local copy as permanent.
Ingest the full corpus safely
Each page in /llms-full.txt begins with a generated header:
# Page Title (/docs/canonical-path)
[Source MDX body]The plugin places --- between pages, but source pages can also contain their own horizontal
rules. Do not split on every ---. Detect generated page headers that match a title followed by a
/docs path, and retain that path as source metadata.
Fetch and validate the response
Require a successful HTTP response and a text content type before replacing an existing corpus.
Detect page boundaries
Start a new document at headers shaped like # Title (/docs/path). Keep the title and path
separate from the MDX body.
Chunk within each page
Split long pages by their Markdown headings while preserving the canonical page path on every chunk. Keep code blocks intact.
Index the API contract separately
Parse /api/openapi.yaml by operation and schema. Do not infer endpoint shapes from examples
or from an APIPage component tag in the MDX corpus.
Use MCP when the agent needs tools
The LLM artifacts provide documentation context only. Use the Y2 MCP integration when an agent also needs bounded report or News reads, OpenAPI operation lookup, or Agent Y2 access.
Separate distribution boundary
@y2-intel/mcp is a separate package whose implementation is not in this platform repository.
After installation, use the MCP client's tools/list, resources/list, and prompts/list
responses as the source of truth for that installed version.
The MCP guide covers client configuration, scoped credentials, capability
inspection, and troubleshooting. Keep Y2_API_KEY in the MCP process environment, never in a
prompt or tool argument.
Prepare a coding agent
Use the maintained integration prompt when an agent will write Y2 API code. It directs the agent to the documentation index, OpenAPI contract, current base URLs, authentication rules, and plan boundaries.
Copy the maintained Y2 integration prompt into your coding agent before it writes API code.
AI coding-agent workflow
Ground an agent, constrain its scope, and verify the generated integration.
OpenAPI contract
Choose the correct server, validate the contract, and regenerate client types.
MCP setup
Configure and inspect the separately distributed MCP server.
Client guides
Build server-side TypeScript, Python, or command-line integrations from the contract.