AI Connect

MCP setup

MCP is how an AI agent connects to a data source. Paste one web address into your agent, sign in to InvestViews, and the agent can look up asking prices for you. Below are the steps for each agent, the local way, and what costs what.

Setup · the main way

Where to paste the URL

Choose the agent you use. Each one asks you to sign in to InvestViews. You do this once.

Claude on the web (claude.ai) and Claude Desktop

Open Customize, then Connectors. If you do not see Customize, look for Settings, then Connectors.
Click + Add, then Add custom connector.
Name it InvestViews. Paste the URL. Click Add.
Click Connect. Sign in to InvestViews and click Allow.

On a Team or Enterprise plan? The account owner adds the connector first, in Organization settings. Then you open Connectors and click Connect.

Claude Code

Run this in your terminal:

claude mcp add --transport http investviews https://mcp.investviews.ai/mcp

Then type /mcp inside Claude Code and pick investviews. Your browser opens. Sign in to InvestViews and click Allow.

Installed the plugin? You do not need this command. See “Run it on your computer instead” below.

Codex

Run these two lines in your terminal. The second opens your browser to sign in.

codex mcp add investviews --url https://mcp.investviews.ai/mcp
codex mcp login investviews

Or add this to ~/.codex/config.toml, then run the second line above:

[mcp_servers.investviews]
url = "https://mcp.investviews.ai/mcp"

Cursor

Add this to ~/.cursor/mcp.json to use it in all your projects. Or open Customize in Cursor and add a custom server. If the file already has mcpServers, add only the investviews entry inside it.

{
  "mcpServers": {
    "investviews": {
      "url": "https://mcp.investviews.ai/mcp"
    }
  }
}

Cursor then asks you to sign in to the server. Click the sign-in prompt, sign in to InvestViews, and click Allow.

Cursor shows as unverified. That is expected. On the sign-in page, check that the address is www.cursor.com, localhost or anysphere.cursor-mcp.

You stay in control. Connecting gives your agent its own read-only access. You can see it in your account, under API tokens. Remove it there and the agent is disconnected.

Your agent is not listed? Look for a setting called “remote MCP server” or “custom connector”, and give it the URL above.

Another way

Run it on your computer instead

For people who like a program on their own machine. You install the InvestViews app and log in with a token. Most people do not need this.

The InvestViews app and its MCP server

Install the app and log in with a token from API tokens. Then your agent starts the server itself.

brew install Investviews/tap/investviews
investviews auth login --token iv_live_xxxxxxxx

Then add the server to your agent. For Claude Code:

claude mcp add investviews -- investviews mcp

Using the Claude Code plugin? It adds this server itself, so skip this command.

For Codex:

codex mcp add investviews -- investviews mcp

Claude Desktop, Cursor and other agents

Add this to the agent's MCP settings file. For Claude Desktop on a Mac that is ~/Library/Application Support/Claude/claude_desktop_config.json. For Cursor it is ~/.cursor/mcp.json. Then restart the agent.

{
  "mcpServers": {
    "investviews": {
      "command": "investviews",
      "args": ["mcp"]
    }
  }
}

If Claude Desktop cannot start it, put the full path in command. Run which investviews to see it.

Older versions of the app do not have this. Check your version, then upgrade.

investviews version
brew upgrade --cask Investviews/tap/investviews

How it is counted

Finding places is free. Only answers cost.

Each MCP tool says if it is free or metered before your agent uses it. So your agent does not stop to ask for permission it does not need, and it does not spend quota it does not have to. The CLI commands count the same way.

Free

Finding things

find_placeTurn a place name into a place your agent can use.CLI: geo search
browse_placesWalk the map of places, from countries down.CLI: geo browse
place_at_pointName the cell and zones at a point on the map.CLI: geo lookup
coverageSee which markets we serve, and how fresh each is.CLI: coverage
filtersSee every value a filter accepts.
usageSee what your token has spent and has left.CLI: usage

Metered

Answers

market_statsThe figures for the newest complete period. A metered call.CLI: stats current
price_historyThe same figures, period by period. A metered call.CLI: stats history
compare_placesSeveral places side by side. A metered call for each place.

Every metered answer ends with a cost: line, so you always see what it cost. The CLI prints a cost: line too.

cost: 1 metered request (group current, …)

What this data is, and is not

Asking prices, not sold prices. The figures come from what sellers list, not from registered transactions. That is a different number, and often a higher one.
Newest complete period, not today. A period closes before it is published, so the latest figure can be several weeks behind. The period is on every answer.
Cells, not addresses. Figures describe an H3 cell between resolution 4 and 8. A request for a finer cell is refused, never quietly answered with a coarser one.

Documentation

Everything the API can do

Written for people and for agents — every guide has a .md twin, and the whole site is available as one file.

QuickstartGet a token, resolve a place, read your first figure.AuthenticationBearer tokens, read-only vs read/write, and the 401 / 402 / 403 split.QuotaEndpoint groups, the X-Quota headers, why discovery is uncounted.DataPeriods, windows, outlier trimming, estimated cells, coverage.GeographyH3 resolutions, browsing the tree, turning a place into figures.ErrorsError shape, code meanings, retry rules and rate limits.

llms-full.txtEvery guide in one file. Paste it into any agent that takes context.