How to embed branded document generation in an AI agent

What you'll build

By the end of this guide, your AI agent will be able to:

• Accept a request like "Generate a proposal for Acme Corporation"
• Automatically extract Acme's brand from their website
• Generate a professionally branded proposal document
• Return a PDF link and shareable web URL in under 5 seconds total

This works for Claude, ChatGPT, Cursor, and any MCP-compatible agent framework.

Option A: Claude, ChatGPT, or Cursor (zero code)

If you use Claude, ChatGPT Pro/Business, or Cursor, you can add DocRocket as an MCP connector with no code at all:

1. Sign up at app.docrocket.ai (email only, 30-day trial)
2. Add the connector: In your AI's settings, add a custom MCP server at https://mcp.docrocket.ai/v1/mcp-tools/docrocket-control/mcp
3. Sign in via OAuth when prompted
4. Test: "Generate a proposal for Acme Corporation"

See the step-by-step guides: Claude · ChatGPT · Cursor

Option B: Custom agent (LangChain, LlamaIndex, CrewAI, etc.)

For custom agents, connect the DocRocket MCP server programmatically.

Python (MCP SDK)

from mcp import MCPClient
import os

# Obtain OAuth token via your app's auth flow
DR_TOKEN = os.environ["DOCROCKET_OAUTH_TOKEN"]

client = MCPClient(
    "https://mcp.docrocket.ai/v1/mcp-tools/docrocket-control/mcp",
    token=DR_TOKEN,
)

# Extract brand from customer's website (first time only)
brand = client.call("docrocket_brand_from_url",
                   url="https://acme.com")
brand_id = brand["brand_id"]  # store this in your DB

# Generate a branded proposal
doc = client.call("docrocket_generate_document",
                  brand_id=brand_id,
                  template="proposal",
                  content={
                      "title": "Partnership Proposal — Acme Corporation",
                      "sections": [
                          {"heading": "Executive Summary",
                           "body": "..."},
                          {"heading": "Solution Overview",
                           "body": "..."},
                          {"heading": "Pricing",
                           "component": "pricing_table"},
                      ]
                  })

print(doc["web_url"])  # docs.docrocket.ai/d/abc123
print(doc["pdf_url"])  # docs.docrocket.ai/d/abc123.pdf

Best practices

• Cache brand IDs: Store the brand_id in your database after the first extraction. On subsequent document generation calls for that customer, pass the stored ID — no re-crawling needed.
• Re-extract on customer rebrand: If a customer updates their website design, call docrocket_brand_from_url again to get an updated brand profile.
• List templates first: Call docrocket_list_templates once at startup to see all available templates and their required content fields. Pass this list to your AI as context so it knows what it can generate.
• Use components: Define reusable blocks (pricing tables, signature sections) once in app.docrocket.ai. Your agent references them by name — DocRocket handles rendering.

A complete example: sales proposal agent

Here's a more complete pattern for a sales proposal agent that handles the brand caching correctly:

from mcp import MCPClient
from your_db import get_customer, update_customer

client = MCPClient("https://mcp.docrocket.ai/v1/mcp-tools/docrocket-control/mcp",
                   token=DR_TOKEN)

async def generate_proposal(customer_id: str, ai_content: dict) -> dict:
    customer = await get_customer(customer_id)

    # Use cached brand ID or extract for the first time
    if not customer.docrocket_brand_id:
        brand = client.call("docrocket_brand_from_url",
                           url=customer.website_url)
        await update_customer(customer_id,
                              docrocket_brand_id=brand["brand_id"])
        brand_id = brand["brand_id"]
    else:
        brand_id = customer.docrocket_brand_id

    # Generate document — AI has already written the content
    doc = client.call("docrocket_generate_document",
                  brand_id=brand_id,
                  template="proposal",
                  content=ai_content)

    return {"url": doc["web_url"], "pdf": doc["pdf_url"]}