Bankruptcy

What Bankruptcy does

Bankruptcy exposes business bankruptcy docket metadata from US federal bankruptcy courts through MCP. The corpus contains approximately 1.55 million filings and is sourced from CourtListener / Free Law Project. Records can identify that a case exists, the debtor name as recorded by the court, the court, filing date, termination date when recorded, and assigned judge.

This is a business-focused dataset. Consumer bankruptcy cases are excluded using a legal-entity-suffix test applied within bankruptcy courts. As a result, a sole proprietorship filed under an individual's name may not be represented. A search with no matches means that no matching business filing was found; it does not establish that an entity has never filed.

How it works

MCP clients can connect to the hosted streamable HTTP endpoint:

{
  "mcpServers": {
    "bankruptcy": {
      "url": "https://gateway.pipeworx.io/bankruptcy/mcp"
    }
  }
}

The endpoint lists the Bankruptcy tools along with shared Pipeworx gateway tools. ask_pipeworx can route natural-language questions across the wider catalog, while the dedicated tools provide direct access to this pack. A full gateway connection is also available at https://gateway.pipeworx.io/mcp.

For clients without MCP support, the same tools can be called over HTTP. The npm package @pipeworx/mcp-bankruptcy provides a local stdio bridge that forwards requests to the hosted endpoint and requires network access. Anonymous use is capped at 50 calls per day; a free Pipeworx account increases the limit to 200 calls per day with an API key.

Setup and configuration

The hosted connection needs no authentication for initial calls. To use the npm bridge, configure the following entry:

{
  "mcpServers": {
    "bankruptcy": {
      "command": "npx",
      "args": ["-y", "@pipeworx/mcp-bankruptcy"]
    }
  }
}

Set PIPEWORX_API_KEY in the server environment to authenticate the bridge with a Pipeworx account. Set PIPEWORX_META_TOOLS=1 if the local bridge should also list the shared gateway tools. When using an HTTPS proxy, NODE_USE_ENV_PROXY=1 may be needed for Node's built-in fetch to honor proxy settings, depending on the Node release.

Tools and capabilities

• bankruptcy_search finds filings whose court-recorded case names contain a company-name match. Results include the court, relevant dates, and judge.
• bankruptcy_recent returns recent filings in newest-first order and can optionally limit results to one court.
• Court identifiers use CourtListener conventions, including nysb, deb, txsb, cacb, and ilnb.
• HTTP callers can inspect a tool with GET /v1/tools/bankruptcy_search and invoke it with POST /v1/tools/bankruptcy_search.

Limitations and notes

The data is a quarterly snapshot. Each response includes snapshot_date, so filings after that date will not appear. Case names preserve court-recorded abbreviations and inconsistent entity suffixes; substring matching helps accommodate those variations.

Termination dates are populated for 78.3% of the corpus. A missing date_terminated value means that no termination date was recorded, not that the case remains open. nature_of_suit is sparse and should not be treated as a dependable case-type filter.

The dataset contains docket metadata only. It does not include petitions, schedules, motions, orders, or other filing documents; those belong to a separate RECAP corpus. The integration is unofficial and is not affiliated with or endorsed by the upstream data provider. The project is licensed under MIT, while the underlying CourtListener / Free Law Project data is published under the Public Domain Mark.