Important Notice: Managing API Usage with AI-Powered MCP Integrations

When an AI assistant is connected to the Regrid MCP and given access to your API token,
every parcel returned by lookup, query, and data retrieval request it makes counts against your API usage quota. Unintentional or unoptimized prompts can result in significantly higher usage than
expected.

Why This Matters

AI assistants connected via MCP can autonomously make multiple API calls in a single
conversation turn. For example, a single prompt asking to "find all parcels owned by a company
in a county" may trigger:

• An initial count query
• One or more paginated data retrieval requests returning up to 50 records each
• Additional follow-up requests to resolve ambiguities or enrich results
• Batch point lookups enriching entire CSV files in a single session

Each of these calls consumes API credits. Without care, a brief AI-assisted session can consume
the equivalent of hundreds of manual API requests.

Best Practices to Control Usage

1. Use return_count Before Retrieving Data

Before fetching full parcel records, ask the AI to first return the count of matching parcels
using the return_count=true parameter. This is a lightweight request that tells you how many
results exist — before committing to a potentially large, credit-intensive data pull.

GET /api/v2/parcels/query
  ?fields[state2][eq]=TX
  &fields[owner][ilike]=7 ELEVEN
  &return_count=true
  &token=YOUR_TOKEN

Response:

{ "count": 313 }

Once you know the count, you can make an informed decision about whether and how to retrieve
the full dataset.

Example prompt:
"How many parcels are owned by Acme Corp in Harris County TX? Return the count only."

2. Use the limit Parameter to Control Result Size

The limit parameter controls the maximum number of parcel records returned in a single API
response (up to 50 per request). When working with an AI assistant, always instruct it to use
an appropriate limit rather than defaulting to the maximum.

This is especially important when:

• Previewing data — set limit=1 to verify the response structure before pulling more
• Iterating in stages — retrieve a small batch, confirm quality, then continue
• Working with large datasets — avoid pulling all records at once by paginating with
  controlled limits and reviewing each page

GET /api/v2/parcels/query
  ?fields[geoid][eq]=48113
  &fields[owner][ilike]=7 ELEVEN
  &limit=5
  &token=YOUR_TOKEN

Example prompt:
"Return only the first 5 parcels owned by 7-Eleven in Dallas County. I'll review those before
deciding how many more to retrieve."

3. Request One Record at a Time When Exploring

When exploring or testing, instruct the AI to return a single result first. Confirm the data
looks as expected before asking for additional records. This prevents large, unnecessary data
pulls during discovery.

Example prompt:
"Look up the first parcel only and show me the result before continuing."

4. Be Specific in Your Prompts

Vague prompts can cause the AI to make broad, open-ended queries that return far more data than
needed. Be specific about geography (county, ZIP, city), owner name format, and the fields you
actually need. The more precise your prompt, the more targeted — and efficient — the API call.

5. Keep Your API Token Secure

Only share your API token in trusted, private AI sessions. Your token authorizes API calls on
your behalf — if exposed in a shared environment or public chat, it could result in unauthorized
usage charged to your account.

Quick Reference: Prompting the AI Efficiently

Parameter Reference

Need Help?

If you have questions about your usage or quota with the Regrid MCP, contact our support team:

• 📧 Support portal: [email protected]
• 📖 API documentation: support.regrid.com
• 🔑 Manage your token: app.regrid.com