The full upstream README, mirrored here for reference. Install config, tool schemas, adoption signals, and an original overview live on the CourtAPI — Federal Court Data listing page.
Search and retrieve US federal court cases, dockets, claims, and documents via PACER — directly from Claude and other MCP-compatible AI assistants.
This MCP server exposes CourtAPI as a set of tools that let an AI assistant:
Coverage: all US federal courts — bankruptcy, district, and appellate. State courts are not supported.
CourtAPI account — Get credentials (APP_ID + APP_KEY) at the developer portal. A free 30-day sandbox is available (150 API calls/day, 500/month) — contact support@courtapi.com to start one.
PACER account — Required for fetching live data (docket updates, document downloads). Register free at pacer.uscourts.gov. PACER charges per page for documents; CourtAPI passes these fees through at cost.
Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):
Restart Claude Desktop. The hammer icon will show CourtAPI tools listed.
The server uses stdio transport by default, which is compatible with any MCP host. Use the same command/args/env pattern above.
For clients that use HTTP rather than stdio (e.g. the OpenAI Responses API remote tools):
| Variable | Required | Description |
|---|---|---|
COURTAPI_APP_ID | Yes | Your CourtAPI application ID |
COURTAPI_APP_KEY | Yes | Your CourtAPI application key |
COURTAPI_BASE_URL | No | Override API base URL (default: https://v1.courtapi.com) |
Once connected, try these prompts in Claude:
"Find the Sbarro bankruptcy case and show me the latest docket entries"
"Search for Chapter 11 cases filed in the Delaware bankruptcy court in 2024"
"Get the claims register for case 1:14-bk-10557 in the SDNY bankruptcy court"
"Set my PACER credentials to username johndoe and password mypassword"
find_casesDiscover cases you don't have a court code or full case number for. Pass a party/company name, partial case number, or free-text keywords. The tool automatically picks the best search endpoint:
search-filings) — full-text across CourtAPI's filing repository; costs 1 API call per 25 results returnedReturns court_code + case_number pairs needed by all other tools.
get_caseGet all metadata for a known case: title, chapter, judge, key dates (filed/closed/discharged), assets, liabilities, and the menu object. The menu has one entry per sub-resource (dockets, claims, parties, etc.) with a modified timestamp — null means that resource has never been purchased from PACER.
Always call this first after finding a case to check what data is already available. Free unless include_live_pacer=true.
get_docketsGet the docket sheet — all filings in chronological order with dates, descriptions, and attached document info. Supports keyword search and date/number range filtering.
Two useful fields on every docket entry:
binder.documents — pre-fetched document metadata including is_cached and download_cost for each PDF, so you can check availability without a separate documents API call.annotations — key dates extracted from docket text (deadlines, hearings, etc.) with key_phrase, datetime, and time_zone. Useful for building calendar integrations without parsing docket text.When calling with include_live_pacer=true to buy a fresh docket sheet from PACER, always pass date_filed_from to limit the purchase to entries newer than what CourtAPI already has. Omitting it buys the entire docket history — expensive for active cases.
get_documentGet a PDF attached to a docket entry. Checks the cache first:
is_cached=true or download_cost="0.00" → returns download_url, preview_url, and ocr_link at no costpurchase=false (default) → returns cost info so you can decide before spendingpurchase=true → buys from PACER, stores it, returns download_url + PACER receiptdocket_seq comes from docket entries (e.g. "42.00000"). part defaults to 1 (main document); exhibits and attachments have higher part numbers.
Once purchased, the document is cached — all subsequent GETs are free.
get_claimsGet the bankruptcy claims register — all proofs of claim filed by creditors, with amounts (filed/allowed/paid), claim type, creditor names, and filing dates. Essential for Chapter 11 creditor analysis.
Filter by claim_type (secured, unsecured, priority, admin, etc.) and sort by claim number, filing date, or amendment date.
get_partiesGet all parties in a case (debtors, creditors, plaintiffs, defendants, trustees) and optionally their attorneys with firm name, contact info, and bar number. Attorney records return null (not an error) when they haven't been purchased for this case — use include_attorneys=false to suppress the attorney fetch.
Useful for conflicts checks, service lists, and identifying who is involved in a case.
get_creditorsGet scheduled creditors from bankruptcy Schedules D/E/F. This is different from get_claims — creditors listed here may not have filed a proof of claim. Use both to get a complete creditor picture.
ncl_searchSearch PACER's National Case Locator across all federal courts at once. More comprehensive than find_cases for party name searches because it queries PACER directly. Costs PACER credits.
Supports: name searches, SSN/TIN lookups (with last name), case number, chapter, court, and date range filters. Paginate with search_id from the previous response.
list_courtsList all ~200 PACER courts with their court codes and names. Filter by type: bankruptcy, district, or appellate.
Common codes: nysbke (NY Southern Bankruptcy), debke (Delaware Bankruptcy), ilnbke (IL Northern Bankruptcy), cacdbe (CA Central Bankruptcy), txnbke (TX Northern Bankruptcy).
get_pacer_credentialsManage PACER credentials stored in CourtAPI. PACER credentials are required before any include_live_pacer=true call.
action | What it does |
|---|---|
check | Show the stored PACER username (password not returned) |
set | Store pacer_user + pacer_pass (validates against PACER by default) |
validate | Test credentials against PACER without storing |
delete | Remove stored credentials |
CourtAPI has two independent cost layers:
CourtAPI API credits — charged by your CourtAPI plan. Free for: case GET lookups, party/title searches, docket GETs from cache, document GETs from cache. Costs 1 call per 25 results for keyword search (search-filings).
PACER fees — charged by the US federal courts per page. Only incurred when:
include_live_pacer=true on dockets, claims, creditors, or case refreshpurchase=true on get_documentdownload_cost="0.00" means a document is cached and free to retrieve. Always check before purchasing.
CourtAPI requires long-form PACER case numbers. PACER often shows short forms in its UI (e.g. 14-10557) but the API always needs the full form.
Non-appellate courts (district, bankruptcy):
O — office/division number (e.g. 1, 2, 3)YY — two-digit filing yearTT — case type: bk bankruptcy, cv civil, cr criminal, ap adversary proceedingNNNNN — sequence numberExample: 1:14-bk-10557 (not 14-10557)
Appellate courts:
Example: 23-1234
If PACER shows a short form, use find_cases to resolve it. Always pair a case number with its court_code — the same sequence number can exist in multiple courts.
When get_dockets or get_claims fetches live data from PACER (include_live_pacer=true), the response includes links.screenshot.pdf.href — a link to a PDF screenshot of the actual PACER page that was fetched. Useful for audit and compliance verification.
MIT — © CourtDrive. See courtapi.com for API terms.