Federal civil litigation

Federal civil litigation in US district courts — everything not captioned United States v. .... The largest categories are prisoner civil rights and habeas petitions, then civil rights employment, product liability and contract.

Part of Pipeworx — an MCP gateway connecting AI agents to 1864+ live data sources. This is an independent, unofficial integration — not affiliated with, endorsed by, or published by the upstream provider.

No API key.

What this is, and what it is not

Docket metadata: that a case exists, its parties, court, docket number, filing and termination dates, and assigned judge. It contains no filings — no complaints, motions, briefs or orders. Those are RECAP, a separate corpus that is not keyless.

An agent with this data knows a case was brought. It does not know what anyone argued. One narrow exception, below: the TEXT of a docket entry's description is searchable without a token even though the documents themselves are not, and that is what makes removal detectable.

Related MCP server: federal-appeals

Removed cases: removed_from

A case removed from state to federal court is two records, and nothing on the federal one says so. njd 2:26-cv-05788 is NS Pharma's removal of New Jersey Superior Court BER-C-000117-26, and its cause reads 09:1 U.S. Arbitration Act — the federal statute invoked, not where the case came from. Both tools therefore look up the originating court by default and return it on the row:

"removed_from": {
  "state_court": "Superior Court of New Jersey, Chancery Division: Bergen County",
  "state_docket_number": "BER-C-000117-26",
  "notice_filed": "2026-05-20",
  "notice_url": "https://www.courtlistener.com/docket/73371682/1/capricor-therapeutics-inc-v-nippon-shinyaku-co-ltd/",
  "source": "Notice of Removal filed on the federal docket (CourtListener RECAP filing text)"
}

Its absence is not a negative. No removed_from means no Notice of Removal was found, and the filing text comes from RECAP, which covers only dockets somebody has purchased or uploaded. Every response that ran the lookup says so in removal_note, and removal_lookup carries the counts — including how many notices were found but carried no parseable court and number, which are reported rather than emitted as half a field. A throttled lookup surfaces as removal_lookup.error, so "could not ask" never reads as "not removed".

The clause is often not there, and that is the real limit. The lookup finds the Notice of Removal reliably; whether that notice's text names the state court is up to whoever docketed it. Measured live on 2026-10-11 across four dockets carrying a notice, ONE wrote the parseable form (... from Superior Court of New Jersey, Chancery Division: Bergen County, case number BER-C-000117-26) and three did not — two read only NOTICE OF REMOVAL by <party> (Filing fee ...) and one had no description at all. A small sample, so take it as "expect misses", not as a rate. Those three are counted in removal_lookup.notices_unparsed rather than having a court inferred from, say, an attachment titled "PCCP Court Docket" — that inference is exactly the plausible-but-fabricated link this field exists to avoid.

A structured originating-court field does exist on CourtListener's /api/rest/v4/dockets/ endpoint and would not have this limit, but it answers HTTP 401: it needs a Free Law Project account token. See the pack header for that trade-off.

Set check_removal: false to skip it and save one upstream request.

Tools

Caveats

• Quarterly snapshot. Every response carries snapshot_date.

• date_terminated is 78.3% filled, so a missing date means not recorded, never still open. Responses say which.

• nature_of_suit is sparse — 14.8% corpus-wide, 30.5% in district courts. This can tell you a company is in federal court; it often cannot tell you what for.

• Party search matches a substring of the case caption, so a distinctive party name works better than a full caption.

• The removal lookup is capped at 20 dockets per call, in chunks of 10, because the upstream filing search is throttled to 5 requests/minute and about 125/day for a token-less caller. When a response returns more rows than that, removal_lookup.not_checked says how many were left unexamined — treat those as unknown rather than as not removed.

Attribution

Data is CourtListener / Free Law Project, published under the Public Domain Mark. Free Law Project funds and maintains the corpus. If this pack is useful to you, https://free.law/donate/ is where that goes.

Quick Start

Add to your MCP client (Claude Desktop, Cursor, Windsurf, etc.):

{
  "mcpServers": {
    "federal-civil": {
      "url": "https://gateway.pipeworx.io/federal-civil/mcp"
    }
  }
}

What this endpoint actually serves

tools/list at https://gateway.pipeworx.io/federal-civil/mcp returns the tools in the table above plus the shared Pipeworx meta-tools — ask_pipeworx, discover_tools, search_within, remember/recall and the rest of the gateway-wide set. So the tool count you see is larger than this table: a single-pack endpoint currently lists roughly 30 shared tools alongside the pack's own. The connection's initialize response states its exact scope, and is the authoritative answer for a given day.

This is deliberate, not multiplexing by accident. The meta-tools are what let a scoped connection answer a question this pack does not cover — via ask_pipeworx, which routes across the whole catalog — without you adding a second MCP server. There is currently no way to mount a pack endpoint without them; if the extra schemas cost you more context than the routing is worth, connect to the full gateway once rather than to several pack endpoints.

Or connect to the full Pipeworx gateway to get every pack's tools listed directly, instead of just this one's:

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

Both URLs reach the same gateway and the same 1864+ data sources. The only difference is which pack's tools are listed directly; ask_pipeworx reaches all of them from either one.

No MCP client? Call it over HTTP

curl -X POST https://gateway.pipeworx.io/v1/tools/federal_civil_search \
  -H 'Content-Type: application/json' \
  -d '{"party":"Monsanto","limit":5}'

No account needed for the first calls. Inspect any tool: GET https://gateway.pipeworx.io/v1/tools/federal_civil_search. Find one: POST https://gateway.pipeworx.io/v1/tools/search_packs with {"query":"..."}.

Run it from npm

@pipeworx/mcp-federal-civil is a local stdio bridge to the Pipeworx gateway: your MCP client starts it with npx, and it forwards every call to the same https://gateway.pipeworx.io/federal-civil/mcp endpoint shown above. It needs network access; there is no offline mode.

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

Without a key it calls anonymously, capped at 50 calls a day. A free account raises that to 200 a day: get a key at https://pipeworx.io/signup and add "env": { "PIPEWORX_API_KEY": "<your key>" } to the server entry. The key is sent as a Bearer token, the same way a hosted connection sends it, so calls are metered and billed exactly as they are there.

It lists only this pack's tools by default. Set PIPEWORX_META_TOOLS=1 to also list ask_pipeworx and the other gateway-wide tools.

Behind an HTTPS proxy: Node's built-in fetch does not read HTTPS_PROXY by default. Newer Node releases honour it when NODE_USE_ENV_PROXY=1 is set; check your Node version's documentation.

Using with ask_pipeworx

Instead of calling tools directly, you can ask questions in plain English — this works on the pack endpoint above as well as on the full gateway:

ask_pipeworx({ question: "your question about Federal Civil data" })

The gateway picks the right tool and fills the arguments automatically.

More

• Docs and guides

• pipeworx.io

License

MIT