United States Judicial Registry
KnowJudges.com
Menu

KnowJudges MCP

Add judge and court research tools directly to an AI assistant.

The remote server exposes 10 read-only tools over Streamable HTTP: judge and court discovery, profiles, biography, courtroom practice, reported legal-AI cases, recorded changes, and immigration outcomes. Tools advertise input and output schemas and return structured, source-linked public information.

Contract checked against live tools/list on October 12, 2026. All tools declare readOnlyHint: true, destructiveHint: false, and openWorldHint: false. They cannot file documents, email judges, or change records.

Start here

Connect once. Then ask your question.

KnowJudges public access is free and read-only. Your AI app's plan and workspace settings determine whether you can add a custom connector.

Claude

Enable KnowJudges in Claude

  1. Open Customize → Connectors. Choose Add → Custom → Web (or Add custom connector).
  2. Name it KnowJudges. In the server URL field, paste:
    https://mcp.knowjudges.com/mcp
    Choose No sign in if asked, then finish adding it.
  3. In a new chat, open + → Connectors and enable KnowJudges. Then paste the example question below.

For a managed team, an owner may need to add it first. Claude connection guide.

ChatGPT

Enable KnowJudges in ChatGPT

  1. Open ChatGPT Plugins. Choose + → Add custom MCP server.
  2. Name it KnowJudges. In the server URL field, paste:
    https://mcp.knowjudges.com/mcp
    Select No authentication if asked. Review the connection notice and choose Create as a plugin.
  3. Install or enable it, start a new chat, then type @ and select KnowJudges before pasting the example question below.

If custom MCP servers are unavailable, check your account or ask your workspace administrator. Some interfaces use Settings → Apps → Create with developer mode. OpenAI connection guide.

Gemini

Enable KnowJudges in Gemini

  1. Open Gemini on your computer. Go to Settings → Connected Apps (sometimes under Personal Intelligence).
  2. Under Custom apps, choose Add a custom app. In the MCP server URL field, paste:
    https://mcp.knowjudges.com/mcp
    Click Next and follow the setup steps.
  3. In a new chat, type @ and select KnowJudges. Then paste the example question below.

Currently requires a personal US account, age 18+, English and Keep Activity on. KnowJudges requires no sign-in or API key; its Gemini connection has not yet been tested. If setup requires credentials, try another app here. Google connection guide.

Perplexity

Enable KnowJudges in Perplexity

  1. Open Perplexity. Go to Account settings → Connectors → + Custom connector and select Remote.
  2. Name it KnowJudges. In MCP Server URL, paste:
    https://mcp.knowjudges.com/mcp
    Set Authentication to None and Transport to Streamable HTTP.
  3. Review the notice, click Add, then open the connector card to enable it. Select KnowJudges for your query and paste the example question below.

Custom connectors are available on Pro, Max and Enterprise; a workspace administrator may need to allow them. Perplexity connection guide.

Other AI apps

Use an app that supports remote MCP

  1. Open the app's MCP, integrations or connector settings and add a remote server named KnowJudges.
  2. In the server URL field, paste:
    https://mcp.knowjudges.com/mcp
    Choose Streamable HTTP and No authentication if asked. No API key or KnowJudges account is required.
  3. Enable its tools in your chat, then paste the example question below.

An app must support remote MCP to use these tools. A pasted prompt alone cannot install the connection.

Try it

Copy a question into your connected AI

Using KnowJudges, find Amy Totenberg in Georgia and summarize her profile and available courtroom rules. Keep it brief, cite sources, and distinguish judge-specific rules from broader rules.

Replace the judge's name and state with your own. You can also ask for a court's information, documented AI citation cases, or recorded immigration outcomes.

Look for a KnowJudges tool call and source links in the answer. If the AI cannot access KnowJudges, check that the connector is enabled for that chat.

Bounded public research

Caller quotas and a separate network backstop

Anonymous callers use the IP fallback: 20 tool requests per hour, 100 per day, and 50 newly accessed judge/court identities per day. OpenAI requests with _meta["openai/subject"] use 60 per hour, 300 per day, and 100 new identities per day. Changing conversation sessions does not reset that subject quota.

Recognized Claude requests use session context when available (60 per hour, 300 per day, 100 new identities per day), without assuming a durable user identity. Without a session, aggregate network controls apply. Recognized-provider network ceilings are 1,000 requests per hour, 10,000 per day, and 1,000 new identities per day; ordinary networks have 200 per hour, 1,000 per day, and 100 new identities per day. These are backstops, not individual entitlements.

Need a dataset or recurring feed? Contact us about bulk data access.

Connect

For AI clients with remote MCP support

  1. Open the client's connectors, integrations, or MCP settings.
  2. Add a custom remote MCP server at https://mcp.knowjudges.com/mcp.
  3. Choose Streamable HTTP if asked. No authorization header is required.
  4. Discover and enable the tools. If the catalog is stale, refresh it in the client.
{
  "mcpServers": {
    "know-your-judge": {
      "url": "https://mcp.knowjudges.com/mcp"
    }
  }
}

Configuration names vary. Use the URL alone if that is what your client accepts. Live tools/list is authoritative; client catalogs and third-party listings may lag a release.

Available tools

Choose the tool for the question

  • find_judgesResolve a named judge and select the returned profile_url. Named matches may have historical or uncertain service. Required: query (at least 2 characters). Optional: state, court, county, limit (default 5; 1–10).
  • get_judge_profileRetrieve identity, assignments, public profile information, sources, and scoped findings. Required: judge_url from find_judges. Use the biography or practice tool for focused questions; missing fields are unavailable information.
  • get_judge_biographyRetrieve sourced birth details, education, professional career, and judicial service. Required: judge_url. Preserve date precision and sources; age is null unless a valid full birth date is known.
  • get_courtroom_presentation_rulesAnswer a named judge’s practice question with relevant findings and underlying sources, without an unrelated court/roster lookup. Required: judge_url. Optional: topics (up to 10 strings). Broader rules are not personal preferences; conditional guidance remains conditional.
  • find_ai_abuse_casesSearch reported legal-AI error cases by case name, error type, or recorded outcome. Sanctions and no sanctions are distinct. Provide meaningful query text or narrowing filters. Optional: query, state, court, judge, response, from_date, to_date, limit (default 5; 1–10). Dates are inclusive YYYY-MM-DD.
  • get_recent_rule_changesRetrieve recorded profile/rule updates since an inclusive date; court filtering includes associated indexed judge profiles. Required: since (YYYY-MM-DD). Optional: judge_url, court_url, limit (default 10; 1–25). Empty results do not prove no earlier real-world changes.
  • find_courtResolve court identity for court information or roster questions and select the returned court_url. Required: query (at least 2 characters). Optional: jurisdiction (2-letter code, e.g. GA, DC, US), state, limit (default 5; 1–10).
  • get_courtRetrieve one court’s public identity, available metadata, and bounded indexed current-status roster. Required: court_url from find_court. Optional: limit (default 10; 1–10). Preserve roster_coverage, roster_truncated, and location-specific contact notes.
  • get_immigration_analyticsRetrieve recorded outcomes for one immigration judge or court with source, denominator, and period coverage. Required: entity (judge or court), profile_url (canonical public judge or court URL). Optional: metric (asylum_denial_rate by default; proceeding_removal_rate only when requested), year (1950–2100), year_basis (calendar by default, or fiscal), nationality (2–100 characters), state (2-letter code).
  • compare_immigration_analyticsCompare highest/lowest recorded judge or court rates jointly filtered by geography, nationality, and period. Required: entity (judge or court). Optional: metric, year, year_basis, nationality, state as above; order (ascending by default, or descending), minimum_decisions (default 1; 1–100000), limit (default 3; 1–10).

For judge search, state accepts a name or postal code. Court search also accepts a two-letter jurisdiction code. Dates use YYYY-MM-DD; immigration state filters require a two-letter code.

Judge workflow

Chain the returned public URL

find_judges({"query":"Amy Totenberg","state":"GA","limit":3})
→ select the matching results[i].profile_url
→ get_judge_profile({"judge_url": selectedProfileUrl})
  OR get_judge_biography({"judge_url": selectedProfileUrl})
  OR get_courtroom_presentation_rules({"judge_url": selectedProfileUrl})

Use the returned profile_url as judge_url. Pick the follow-up that answers the question; all three calls are not required. Do not construct URLs from names or use internal IDs, keys, or bare slugs. A practice question does not require a court roster lookup.

Court workflow

Preserve court identity and roster coverage

find_court({"query":"U.S. District Court for the Northern District of Georgia","state":"GA"})
→ select the matching results[i].court_url
→ get_court({"court_url": selectedCourtUrl,"limit":5})
→ get_recent_rule_changes({"since":"2026-10-01","court_url": selectedCourtUrl})

Pass the exact returned court_url, including its jurisdiction. Court counts cover indexed active, senior, recalled, or visiting profiles, not an independently verified complete bench. Keep roster_coverage and roster_truncated. Empty means no indexed current matches; a null count means unavailable roster data. Neither establishes that a court has no judges. Preserve location-specific contact notes.

Judge-specific change lookups use judge_url instead. Tracking starts at change_tracking_started_at; an empty change result is not proof no real-world rule changed.

Legal-AI cases

Keep the recorded outcome and link status

find_ai_abuse_cases({
  "query": "AI hallucinations fabricated legal citations",
  "response": "Monetary Sanctions",
  "limit": 5
})

Preserve decision_link_status, dataset_source, and summary_note. Decision links may be verified destinations or labeled source fallbacks; a fallback is not a verified official court decision. Dataset inclusion alone does not establish a judicial finding of AI use. total_matches is not the number of supplied records; unavailable ruling counts are not zero.

Immigration outcomes

Show decision counts alongside rates

compare_immigration_analytics({
  "entity": "judge", "nationality": "Brazil",
  "state": "CA", "year": 2025,
  "order": "ascending", "limit": 3
})
→ lowest recorded asylum denial rates
→ repeat with order: "descending" for highest rates
→ get_immigration_analytics({
    "entity": "judge", "profile_url": returnedProfileUrl,
    "nationality": "Brazil", "state": "CA", "year": 2025
  })

Examples illustrate inputs, not guaranteed coverage for a nationality or year. Use an available, non-null returned public URL as profile_url. Calendar/fiscal years refer to proceeding completion. Asylum percentages use grants plus denials; procedural outcomes are separate. Samples below 35 decisions are labeled, not hidden; zero decisions mean no rate, not 0%.

Keep availability, message, denominator/geography definitions, source, coverage, and limitations. Historical rates do not predict an individual outcome or imply that a person can choose a judge.

Limits and request accounting

Focused queries rather than enumeration

Searches, court rosters, and immigration comparisons return at most 10 records; recent changes return at most 25. Practice filters accept up to 10 topics. No pagination, cursor, offset, batch enumeration, or bulk export is available. Empty, wildcard, one-character, jurisdiction-only judge/court searches, and all-judge/all-court requests are rejected. Provide identifying name/location text plus narrowing filters.

Sequential prefix/identifier harvesting is throttled. Revisiting an already counted judge/court does not consume another daily unique identity unit. Initialization, discovery, and control requests count toward the network backstop, not the normal tool quota. REST shares enforcement and uses anonymous fallback without MCP caller context. Limits can return HTTP 429 or a bounded MCP tool failure. Respect retry information; do not rotate identities to bypass controls.

Sources and verification

Keep evidence and uncertainty attached

Retain public KnowJudges URLs, source/official URLs, source titles, excerpts, scope/applicability, effective/revision dates, and last checked dates where returned. Judge-specific, courtwide, statewide, and nationally applicable federal/EOIR rules have different scopes. Broader rules are not personal preferences; conditional guidance remains conditional. Missing findings establish neither permission nor prohibition.

last_updated describes publication; last_verified or source_checked_at describes an evidence check. A publication update does not mean every source was checked that day. Preserve supported year-only dates and factual uncertainty. An assignment end date alone does not establish retirement or a senior-status transition. Missing data and tool failures are different states. Check the source and case-specific orders before relying on a requirement.

Updated on November 1, 2026.