MCP Server - Crustdata Docs

Quick start

You need a Crustdata account to use the MCP server. Sign in or create one. The endpoint is the same everywhere:

https://install.crustdata.com/mcp

Add to Claude

Opens the Add custom connector dialog on claude.ai.

Add to Cursor

Installs through Cursor’s MCP handler.

Add to VS Code

Registers the server over streamable HTTP.

Add to ChatGPT

Opens the Crustdata listing in the plugin directory.

Claude
Claude Code
Codex CLI
Antigravity
Programmatic access

The Add to Claude card does this in one click. To set it up manually:

Open the connector settings

In Claude, open Customize, then Connectors, then Add custom connector.

Configure the connection

Name: Crustdata
URL: https://install.crustdata.com/mcp

Click Add to save.

Connect

Open the new connector under Not connected and click Connect, then sign in with your Crustdata account.

Connectors are account-wide. Once you add it, Crustdata is available on claude.ai, Claude Desktop, and mobile.

Add the connector from your terminal:

claude mcp add --transport http crustdata https://install.crustdata.com/mcp

Then sign in: run /mcp inside Claude Code and pick crustdata.The Crustdata plugin registers this same connector and adds ready-to-run skills, if you want both in one install.

Add the connector from your terminal:

codex mcp add crustdata --url https://install.crustdata.com/mcp

Codex detects that the server uses OAuth and opens a Crustdata sign-in in your browser. If the sign-in does not open or does not finish, run codex mcp login crustdata.Run /mcp inside Codex to confirm crustdata is listed.The Crustdata plugin for Codex registers this same connector and adds ready-to-run skills, if you want both in one install.

Add Crustdata to Antigravity’s MCP config and sign in with your Crustdata account. No API key needed.Antigravity 2.0, the IDE, and the CLI share one config file, ~/.gemini/config/mcp_config.json. Pick the one you use.

Antigravity 2.0

Open the config folder

In Antigravity 2.0, click Settings at the bottom left, then Customizations. Under Installed MCP Servers, click Open MCP Config. This opens the folder that holds mcp_config.json (~/.gemini/config/mcp_config.json) in your file manager, not the file itself.

Add the server

Open mcp_config.json in a text editor. On a new install the file is empty: paste the block below as is. If the file already has an mcpServers object, add the crustdata entry inside it.

{
  "mcpServers": {
    "crustdata": {
      "serverUrl": "https://install.crustdata.com/mcp"
    }
  }
}

Use serverUrl for remote servers. Some Antigravity versions reject url and httpUrl. Save the file.

Refresh

Click the refresh icon next to Add MCP. crustdata appears with an Authenticate button and an Unauthorized error. That is expected until you sign in.

Click Authenticate and sign in with your Crustdata account in the browser. Copy the authorization code the browser shows, paste it into the field in Antigravity, and click Submit. The dot next to crustdata turns green when you are connected.

Antigravity IDE

Open the raw config

Click … at the top of the agent side panel, then MCP Servers, Manage MCP Servers, View raw config. This opens ~/.gemini/config/mcp_config.json.

Add the server

If the file is empty, paste the block below as is. If it already has an mcpServers object, add the crustdata entry inside it.

{
  "mcpServers": {
    "crustdata": {
      "serverUrl": "https://install.crustdata.com/mcp"
    }
  }
}

Save the file.

Refresh

On the Manage MCP servers page, click Refresh. crustdata appears with Configure and Authenticate buttons and an Unauthorized error. That is expected until you sign in.

On the same page, click Authenticate next to crustdata. Sign in through your browser. If Antigravity asks for an authorization code, paste it and click Submit.

Antigravity CLI

Add the server

Run this in your terminal, not inside an agy session.

agy mcp add crustdata https://install.crustdata.com/mcp

Run agy mcp list to confirm it is listed.

Start agy and run /mcp. Select crustdata with the arrow keys, press Enter for its actions, and choose Authenticate. Open the URL it shows, sign in, then paste the authorization code back into the CLI.

You can use the same server from your own code: agent frameworks, backend jobs, or anywhere the sign-in flow is not practical. Connect any MCP client to the endpoint and pass your Crustdata API key as a bearer token.

Endpoint: https://install.crustdata.com/mcp
Header: Authorization: Bearer YOUR_API_KEY

For MCP clients that take a JSON config with custom headers:

{
  "mcpServers": {
    "crustdata": {
      "type": "streamable-http",
      "url": "https://install.crustdata.com/mcp",
      "headers": { "Authorization": "Bearer YOUR_API_KEY" }
    }
  }
}

When Crustdata can’t answer a request, your assistant may send us a short report through the feedback tool, so we can fix the gap. The tool tells the assistant to leave out names, emails, and other personal details. Your client asks you before the first report. Allow the tool once and your client stops asking.

Check it worked

Ask your assistant:

How many credits do I have?

A number means you are connected. If you get no number:

Restart your client. A mid-session install may not take effect until it does.
Re-run your client’s sign-in step, which is separate from adding the connector.
In Antigravity, check that the entry uses serverUrl, not url, and that you pasted the authorization code after Authenticate.
Watch the setup walkthrough.

Try it

Ask for what you want in plain English:

“Show me AI startups with 100-500 employees”
“Who are the engineering leaders at Anthropic?”
“What jobs does Stripe have open right now?”
“Find the profile for jane.doe@acme.com”
“Get me the email for this person’s profile”

A single request can chain steps. “Find AI startups hiring engineers, then get me the founders’ emails” runs start to finish without you breaking it up.

What you can do

Companies

Find and enrich companies by industry, headcount, location, and funding, with full firmographic profiles.

People and contacts

Search people by title, seniority, company, and skills, then enrich profiles with business emails, personal emails, and phone numbers.

Jobs

Search millions of job listings with filters and breakdowns, or pull a company’s live openings on demand.

Social and web

Track recent social posts by person, company, or keyword, and search and read content from across the web.

Hand it a list instead of one name and it enriches the whole list, splitting the work into batches on its own.

Credits

Every request reports the exact credits it spent, in real time, measured from what the request actually did rather than estimated. Enrichment takes the cheapest correct path, so you do not over-spend by accident. Costs vary by request and plan. See Pricing for per-request costs and Rate limits for limits.

Catalog resources

The server publishes its own usage guide as MCP resources: markdown pages an agent reads directly instead of re-deriving the tool list and schemas each session. Read crustdata://catalog first. It carries the must-know rules and an index of the sections.

Resource	What’s in it
`crustdata://catalog`	Start here. The must-know rules and an index of the sections below
`crustdata://catalog/quickstart`	How to drive the server: writing a script that composes tool calls
`crustdata://catalog/tools`	Every callable tool with its one-line purpose and cost
`crustdata://catalog/query-builder`	Filters, operators, and matching behavior
`crustdata://catalog/results`	Shaping results: `project`, `compact`, `paginate`, `parallelMap`
`crustdata://catalog/fields`	Common filter columns and identifiers per tool
`crustdata://catalog/patterns`	Recipes: search then enrich, identify then enrich, bulk, pagination
`crustdata://catalog/dry-run`	Testing a script for zero credits before running it for real
`crustdata://catalog/ranking`	Bounding result pages server-side and ranking in code
`crustdata://catalog/cost`	Credit pricing and the cheapest-path rules
`crustdata://catalog/full`	The entire guide in one read

Any MCP client can fetch these with a standard resource read: session.read_resource("crustdata://catalog") in the Python SDK, or client.readResource({ uri: "crustdata://catalog" }) in TypeScript.

Setup walkthrough

Still stuck? Watch the full setup from start to finish.

Get started with the Crustdata MCP

Support

For help or feedback, contact support@crustdata.co.

​Quick start

Add to Claude

Add to Cursor

Add to VS Code

Add to ChatGPT

​Check it worked

​Try it

​What you can do

Companies

People and contacts

Jobs

Social and web

​Credits

​Catalog resources

​Setup walkthrough

​Support

Quick start

Check it worked

Try it

What you can do

Credits

Catalog resources

Setup walkthrough

Support