Skip to content
San DiegoProsperity Observatory

CONNECT · PUBLIC & READ-ONLY

Two minutes to cited answers
in the assistant you already use.

Paste one URL into Claude, ChatGPT, Cursor or any client that supports remote MCP servers. Your assistant gets six research tools and returns records with their original source, locator and record ID. No account, no password, nothing to install.

Streamable HTTP · public read-only research data

Your connection URL

Loading…

Checking the MCP service…

Claude (web or desktop)

  1. Open Settings → Connectors and choose Add custom connector.
  2. Name it “San Diego Prosperity Observatory” and paste the URL above.
  3. Leave authentication empty and save. In a new chat, ask a question and the tools appear automatically.

Claude Code (terminal)

  1. Run this once:
Loading…

ChatGPT, Cursor and other MCP clients

  1. Open the client’s connectors, MCP or tools setting.
  2. Add a remote server. Transport is Streamable HTTP; paste the URL above.
  3. Choose no authentication. The six tools listed below appear by name.

What we log: research questions and search terms sent to the tools are stored privately to find useful datasets and coverage gaps. We do not collect your chat history, account identity or IP address. Keep confidential sources and personal information out of submitted questions.

01 / TRY THESE FIRST

Questions that work today.

Each one is answerable from the reviewed exports and returns rows with citations. The home page shows the actual answers.

“Who is on the November 2026 ballot in Oceanside? Cite the source document and page for each candidate.”

“Compare rent burden, median household income and median gross rent for Encinitas, Vista and San Diego County from the 2020-2024 ACS, with margins of error.”

“List behested payments reported with San Diego Gas & Electric as the payor, under every spelling, with the source row for each.”

02 / A PRACTICAL WORKFLOW

Ask, discover, verify.

  1. Start with a research question.

    Your assistant can call research_question to describe the civic question. That also tells us which sources people need next.

  2. Choose the right source.

    list_datasets and describe_dataset show reporting periods, geography, limitations and reuse conditions before you rely on a file.

  3. Inspect records and evidence.

    search_records and get_record return rows with provenance. Keep the citation and record ID in your notes; open the original before you publish.

03 / THE TOOLS

A small research interface.

MCP is the Model Context Protocol, a standard connection through which assistants request data and evidence. Your assistant writes the answer; the tools supply the records. Verify the cited records yourself.

ToolWhat it does
list_datasetsFind the reviewed, downloadable exports by topic.
describe_datasetRead one export’s scope, dates, limitations and checksum.
search_recordsSearch one export with bounded pagination and provenance on every row.
get_recordRetrieve an exact source record and its original evidence.
get_coverageSee what is in the database, what is open to the public connection and what is staged.
research_questionPrivately log a question and get back the datasets that may answer it.

The public connection searches the reviewed exports listed on the Sources page. The larger database inventory is being released source by source; get_coverage shows the current boundary. Record tools are read-only. The question log writes only service-improvement questions and cannot change civic records.