# CFPB consumer complaints [Health: Active]

**Category:** 💻 Developer Tools  
**Repository:** https://github.com/bnovarini/cfpb-complaints-analysis  
**GitHub Stars:** 0  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/cfpb-consumer-complaints

## Description
CFPB consumer complaints: 18M records, 3.85M archived narratives. Search, counts, trends.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `uvx` (confidence: high):

```json
"mcpServers": {
  "cfpb-consumer-complaints": {
    "command": "uvx",
    "args": ["cfpb-complaints-analysis"],
    "env": {
      "CFPB_DATA_URL": "",
      "CFPB_DATA_DIR": ""
    }
  }
}
```

**Requires environment variables:** `CFPB_DATA_URL`, `CFPB_DATA_DIR` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation

## What CFPB consumer complaints MCP server does

The CFPB consumer complaints MCP server makes CFPB’s public complaint data available to AI assistants through MCP. Its dataset contains 18,239,728 structured complaints received from December 2011 through October 7, 2026, plus 3,851,415 archived narratives published through August 14, 2026. The structured complaint file and narrative archive come from CFPB sources rather than a third-party dataset.

The server supports company lookup, complaint totals, time-series analysis, company profiles, comparisons, narrative search, narrative counts, and individual complaint retrieval. It joins archived narratives to current complaint records using Complaint ID.

## How it works

Install the Python package and run `cfpb-complaints-mcp` as a stdio MCP server:

```bash
pip install cfpb-complaints-analysis
cfpb-complaints-mcp
```

On its first run, the process downloads approximately 700 MB of Parquet data to `~/.cache/cfpb-complaints-analysis`. It can then be added to an MCP client as a local command. A hosted endpoint is also registered as `io.github.bnovarini/cfpb-complaints-analysis`.

The available tools are `dataset_info`, `list_values`, `find_company`, `complaint_counts`, `trend`, `company_profile`, `compare_companies`, `search_narratives`, `count_narratives`, and `get_complaint`. Narrative searches scan text; broad searches typically take about ten seconds, while all-year counts can take 30 to 45 seconds. Supplying a date, company, or product filter improves response time.

## Setup and configuration

Use `CFPB_DATA_URL` when the Parquet files should be fetched from an alternate location. Set `CFPB_DATA_DIR` to point the server at data built or stored locally. The repository also provides a command for building the data from CFPB sources: `cfpb-complaints --data data build`.

The CFPB consumer complaints MCP server can be configured as a local stdio process or accessed through the hosted registry endpoint. The README does not specify credentials or other required environment variables.

## Limitations and notes

Archived narratives do not cover every complaint. CFPB required consumer consent, did not publish narratives before 2015, and stopped publishing new narratives in September 2026. Complaints after the archive cutoff generally have no narrative, and the archive becomes especially sparse near the end of 2026. The live complaint file continues to update when the dataset is rebuilt.

Narratives are scrubbed by CFPB, with personal information represented as `XXXX`. Complaint records are unverified consumer allegations, not confirmed findings. Counts are raw totals and are not adjusted for company size, customer counts, or account volume. A company may appear under multiple names, so use `find_company` and `company_contains` when investigating name variants. Recent months may be partial because complaints can still be sent to companies, and an “In progress” status does not indicate an outcome.

The CFPB consumer complaints MCP server is released under the MIT license. The data itself is CFPB data, and the repository documents reconciliation checks against CFPB’s published figures.

_Full upstream README: https://allmcps.com/mcp/cfpb-consumer-complaints/readme_