> ## Documentation Index
> Fetch the complete documentation index at: https://00doc-web.vercel.app/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Agent context

> Use Unfurl URLs as live context for Claude, Cursor, and other agents — not stale file uploads.

# Agent context

Agents work best when they read **accurate, current** information. Uploading a PDF snapshot freezes the world at upload time. A Unfurl link — or `get_document` over MCP — tracks the **live** document.

## The stale-snapshot problem

When you attach a file to a chat:

* The model sees **that bytes**, not next week's edit.
* You must **re-upload** after every change.
* Large PDFs burn tokens and lose structure.

When you give an agent a **`document_id`** or share URL in a workflow wired to Unfurl:

* `get_document` returns **today's** HTML when you need content.
* `publish_document` **writes** the next version without leaving the tool loop.

## Where to use it

* **Claude Projects** — paste the canonical doc link in project instructions; combine with MCP for read/write.
* **Cursor** — reference the doc in prompts when refactoring specs or ADRs stored on Unfurl.
* **Automation** — a job can poll or trigger publishes so agents and humans never diverge on "source of truth."

## Tie-in to MCP

Closing the loop:

1. Agent reads with `get_document` (`include_html: true` when it must reason over body content).
2. Agent edits in conversation.
3. Agent writes with `publish_document` using the same `document_id`.

<CardGroup cols={2}>
  <Card title="Agent-ready documents" icon="sparkles" href="/core-concepts/agent-ready-documents" />

  <Card title="get_document" icon="download" href="/api-reference/get-document" />
</CardGroup>