# Causilo

> Causilo is an MCP server for analytics and data. Predicts a missing table column from known rows, for Claude and ChatGPT users. It runs as a hosted remote server, so there is nothing to install locally.

- **Kind:** MCP server
- **Platforms:** API, CLI, Web
- **Website:** https://docs.nums.world/assistants
- **MCP Registry name:** io.github.nums-ai/causilo
- **Categories:** [analytics and data](https://ai-powered.com/categories/analytics-and-data)
- **Listing updated:** 2026-11-18

## About

Causilo is a remote MCP server from NUMS that lets an AI assistant run tabular predictions on your data. You give it a table with known values in one column and a second set of rows with that column empty, and it returns predictions for the empty rows. The server chooses between classification and regression from the target column unless you specify one.

The connection is made through a sign-in flow: the assistant opens a Causilo consent page, you confirm a workspace, and Causilo issues a key named after the assistant. The assistant holds only a token accepted by the MCP server, not the key itself. Only prediction calls draw on your allowance, while inspecting, estimating and checking usage are free.

The README and docs describe two routes for larger work. Claude Code and Codex can send a CSV, TSV, Excel or Parquet file through a generated shell command, saving predicted rows to a file. Anything beyond the assistant limits can be sent directly to the Causilo API.

## Features

- **Classification and regression:** Predicts a column of whole-number or categorical targets as classification, and continuous targets as regression.
- **Assistant connection:** Connects to Claude on the web and Claude Code, and to ChatGPT on the web and Codex, using a standard MCP endpoint.
- **Table inspection and estimates:** Shows columns, types, missing values and first rows, and estimates whether a table fits and how long it takes.
- **File-based prediction:** Predicts the empty rows of a CSV, TSV, Excel or Parquet file from the command line in Claude Code or Codex.
- **Usage reporting:** Shows the current allowance and request limits before running predictions.

## Good for

- Filling in missing values in a spreadsheet column using rows where the answer is known
- Estimating a numeric figure such as a price or quantity for new rows in an assistant conversation
- Predicting a category label for a batch of records in a CSV file from the shell
- Checking whether a table is within size limits before sending it for prediction

## Install

### Claude Code (Terminal)

```sh
claude mcp add --transport http causilo https://mcp.nums.world/mcp
```

### Claude Desktop (Settings → Connectors → Add custom connector)

```sh
https://mcp.nums.world/mcp
```

### Cursor (~/.cursor/mcp.json)

```json
{
  "mcpServers": {
    "causilo": {
      "url": "https://mcp.nums.world/mcp"
    }
  }
}
```

### VS Code (.vscode/mcp.json)

```json
{
  "servers": {
    "causilo": {
      "type": "http",
      "url": "https://mcp.nums.world/mcp"
    }
  }
}
```

### Windsurf (~/.codeium/windsurf/mcp_config.json)

```json
{
  "mcpServers": {
    "causilo": {
      "serverUrl": "https://mcp.nums.world/mcp"
    }
  }
}
```

### Grok Bot (Grok Bot desktop app → Connect apps → Add a custom connector)

```sh
https://mcp.nums.world/mcp
```

Enter a Name (for example "Causilo"), choose Type: Remote HTTPS, paste this as the Server URL and choose Add plugin. When we checked, this server asked for OAuth sign-in. If its docs give you an OAuth client ID and secret, enter them in the form's OAuth Client ID and Client Secret fields.

### Other clients (your client's MCP settings (most use this shape))

```json
{
  "mcpServers": {
    "causilo": {
      "type": "http",
      "url": "https://mcp.nums.world/mcp"
    }
  }
}
```

## Tools

- `predict_table`: Predicts one column and returns the predictions.
- `estimate_table`: Checks whether a table fits and roughly how long it takes.
- `inspect_table`: Shows the columns, types, missing values and first rows.
- `usage`: Shows the allowance and the request limits.
- `get_upload_command`: Gives a shell command that predicts from a file. Claude Code and Codex.

## FAQ

### Which assistants can connect to Causilo?

The docs describe connecting Claude on the web and in Claude Code, and ChatGPT on the web and in Codex, by adding the server at https://mcp.nums.world/mcp.

### Which actions use my allowance?

Only a predict_table call or a file prediction when its command runs. Inspecting, estimating and checking usage are free, as is getting an upload command.

### How large a table can be predicted in a chat?

Claude sends up to 1,000 rows in the request, and the answer holds up to 1,000 predicted values per call. Larger tables go through a file in Claude Code or Codex, or through the API.

### Does the MCP server store my data?

The docs say the server holds the table and predictions in memory during the request and does not store them. Disconnecting is done by revoking the connection key in the Causilo console.

### How do I add Causilo to Claude Code?

Run this command in your terminal: claude mcp add --transport http causilo https://mcp.nums.world/mcp

### What are the best alternatives to Causilo?

The closest alternatives to Causilo are aidnn, Data Cleaner, nextanalytica, Laplace Analytics and GetQueryly — AI Data Analyst.

## Where this comes from

- Description: written from the project's website and its MCP Registry entry.
- Status: checked automatically by AIPoweredBot on 19 November 2026.
- Install steps: from the official MCP Registry entry.

## Alternatives

- [aidnn](https://ai-powered.com/mcp/aidnn): aidnn is an MCP server for analytics and data.
- [Data Cleaner](https://ai-powered.com/mcp/data-cleaner): Data Cleaner is an MCP server for analytics and data.
- [nextanalytica](https://ai-powered.com/mcp/nextanalytica): nextanalytica is an MCP server for analytics and data.
- [Laplace Analytics](https://ai-powered.com/mcp/laplace-analytics): Laplace Analytics is an MCP server for e-commerce.
- [GetQueryly — AI Data Analyst](https://ai-powered.com/mcp/getqueryly-ai-data-analyst): GetQueryly — AI Data Analyst is a freemium MCP server for analytics and data.
- [Mycorr](https://ai-powered.com/mcp/mycorr): Mycorr is a freemium MCP server for databases.
- [Europa](https://ai-powered.com/mcp/europa): Europa is an open-source MCP server for analytics and data.
- [PdfParse](https://ai-powered.com/mcp/pdfparse): PdfParse is a freemium MCP server for files and documents.

---
Source: https://ai-powered.com/mcp/causilo
