CourtListener provides searchable access to U.S. case law, court opinions, dockets, judicial decisions, and legal filings. Find cases by citation, keyword, court, or judge; retrieve opinion texts and docket records; and research legal precedents using data from federal and state courts.
Hermes is Nous Research's command-line agent. MCP servers are declared in its YAML config and reloaded without a restart.
This guide connects CourtListener MCP to Hermes through MewCP, so Hermes can use CourtListener MCP's 19 tools, like search, get_endpoint_schema and call_endpoint, without your CourtListener MCP credentials ever reaching the app.
CourtListener MCP signs in through managed OAuth. MewCP stores the token encrypted, refreshes it before it expires and injects it only at call time. Your agent never holds it, so it can never leak it.
More apps
Manage courses, assignments, rosters and student activity within Google Classroom. Create learning workflows, track submissions and organize educational content across classrooms.
AI's impact on real-world sectors, as a structured record for agents: source-linked headlines placed in one of 18 sectors (healthcare, law, retail, government, defense and more), with outlet, country and time, since June 2026
Clerk enables AI agents to manage application authentication, user accounts, organizations, sessions, and access permissions, including inspecting user profiles and configuring authentication workflows.
ConsentStack automates website cookie consent management by creating and publishing consent banners, configuring privacy settings, detecting and categorizing tracking scripts, running compliance scans, and reviewing consent analytics and logs for GDPR and other privacy regulations.
Other clients
CourtListener MCP works the same way in every client below. Pick yours for its own guide.
Free to start.
Scope access to the exact CourtListener MCP tools and accounts you allow, work and personal side by side. Anything outside that scope is refused at call time, even when the model asks for it.
Each call is logged: what ran, on which account, and whether it worked. Failures come back structured, so the agent knows whether to retry instead of guessing.
Every app sits behind one secure MCP URL, and there is no server for you to deploy or keep running. Add an integration in the dashboard and it is live on the next call, with no config to change and no new credentials to wire up. Tools load only when a task needs them, so your agent stays fast and focused however many apps you connect.
Edit config.yaml in your Hermes config directory (~/.hermes).
Put this under the top-level "mcp_servers:" key.
Run /reload-mcp in Hermes, or start a fresh session. Sign in when the browser window opens.
Start a new chat and ask: "List the tools you can use from MewCP." You should see your connected apps.
Search for case law, dockets, judges, and oral arguments. `absolute_url` is a relative path; prefix it with `https://www.courtlistener.com`. Pass IDs, not URLs, to other tools.
Get the schema for a CourtListener API endpoint. Use this for additional API endpoints which do not have a dedicated MCP tool. These endpoints are distinct from the search endpoint and often include more detailed metadata.
Call CourtListener API endpoint. Use this for additional API endpoints which do not have a dedicated MCP tool. These endpoints are distinct from the search endpoint and often include more detailed metadata.
Get an item by ID from a CourtListener API endpoint.
List every valid value for a field whose choices are too long to include in its schema. Only the fields named in `field_name` need this; every other choice field lists its values in the schema where it appears.
Get the number of results from a previous query. Some endpoints return the count lazily. Use this tool to retrieve the count from a previous query if it is not available.
Get more results from a previous query. Use this tool to continue paginating through results returned by the `search` or `call_endpoint` tools.
Read the full text of a court opinion or RECAP document. Fetches the document text and either returns it in full or as one or more paginated chunks. For opinions, uses the ``html_with_citations`` field (the most complete text representation). For RECAP documents, uses ``plain_text``. **Usage patterns** - *Full read*: omit ``chunk_index`` to receive the entire document along with its ``total_chars``. - *Single chunk*: pass an integer ``chunk_index`` (0-based) to receive one window of ``chunk_size`` characters. The response includes ``total_chunks`` so you can jump directly to any part of the document (e.g. set ``chunk_index`` to ``total_chunks - 1`` to read the conclusion). - *Multiple chunks*: pass a list of chunk indexes to retrieve several non-contiguous windows in a single call (up to 10). Useful when you already know the chunk size from a previous call and want the next N pages at once. Input should include exactly one of opinion_id, recap_document_id, or cluster_id.
Search for snippets within one or more court opinions or RECAP documents. Performs a case-insensitive literal search (similar to grep) and returns up to 20 matching excerpts per document with surrounding context. Use this to locate specific language—a party name, a statutory citation, a key phrase—without reading whole documents. Pass a list of IDs (up to 10) to search several documents in a single call. Results are returned as a list; errors on individual documents are included as an ``error`` field so one unavailable document does not abort the rest. When ``match_count`` exceeds ``shown``, the first 20 matches are returned. Use ``read_document`` with ``chunk_index`` to read the area around a match's ``position`` if you need more context. Input should include exactly one of opinion_id, recap_document_id, or cluster_id.
Extract and resolve legal citations from text using eyecite. Runs locally with no API calls or rate limits. Handles all citation types: full case citations, id., supra, short cites, and statutes. Use this tool to understand the citation structure of a document before selectively verifying citations with analyze_citations.
Analyze and verify legal citations against CourtListener. Extracts all citations locally using eyecite, then verifies each unique case citation against CourtListener's database via the citation-lookup API. Returns case name, date, citation count, and verification status for each citation. For documents with more than 250 unique case citations, the first batch is verified immediately and a job_id is returned. Use resume_citation_analysis to continue verifying remaining citations. Terminology in the output: * **citation occurrence** — each citation as it appears in the text. * **unique citation string** — distinct ``volume reporter page`` triples (e.g., one case cited three times is one string). * **unique case cluster** — distinct CourtListener case clusters after parallel-citation dedup (several strings may map to one). Case-name cross-check: when a citation verifies by reporter but its input case name differs significantly from the cluster's canonical name, a WARNING is emitted flagging a possible hallucinated citation. Input should include exactly one of opinion_id or cluster_id.
Resume verifying citations from a previous analysis. Use this after analyze_citations returns with pending citations due to rate limiting (more than 250 unique case citations). Takes the job_id and verifies the next batch.
Create a search alert on CourtListener. Use `call_endpoint` with endpoint_id "alerts" to list existing alerts.
Delete a search alert on CourtListener. Use `call_endpoint` with endpoint_id "alerts" to list existing alerts.
Subscribe to alerts for a docket on CourtListener. Use `call_endpoint` with endpoint_id "docket-alerts" to list existing subscriptions.
Unsubscribe from alerts for a docket on CourtListener. Looks up the alert by docket ID and deletes it, so you only need the docket ID (not the alert ID). Use `call_endpoint` with endpoint_id "docket-alerts" to list existing subscriptions.
Request a RECAP document that is not yet available on CourtListener through Pray and Pay. Prayers are pooled across users: when anyone buys the document from PACER, it is added to CourtListener for free and everyone praying for it is emailed. Nothing is charged to the user, but fulfillment is not immediate and not guaranteed. Only documents whose `is_available` is false need a prayer. This tool checks first and, if the document is already available, tells you to read it with `read_document` instead. Prayers count against a daily per-user limit, so only pray when the user actually wants the filing. Use `call_endpoint` with endpoint_id "prayers" to list the user's pending prayers.
Withdraw a pending Pray and Pay request for a RECAP document. Looks up the user's pending prayer by RECAP document ID and deletes it, so you only need the document ID (not the prayer ID). Prayers that have already been granted cannot be withdrawn and do not need to be: the document is available. Use `call_endpoint` with endpoint_id "prayers" to list the user's pending prayers.
Check the user's CourtListener API usage and rate limits. Returns live usage per throttle scope, daily request counts for the last 14 days, and membership status. The `user` scope is the one that matters: it is the main API quota and governs nearly every endpoint and tool. When the user asks about their rate limit, remaining requests, or 429 errors, answer from the `user` scope; `summary` states it in plain language. The other scopes are narrow: `citations` counts citation lookups, `fetch` counts PACER fetch requests, and `api_usage` only limits this usage check itself. Do not confuse `api_usage` with the user's API usage. Each limit reports `used`, `limit`, `remaining`, `reset_at` (null when the window is empty) and `blocked`. This tool never spends the `user` quota, so it works while other tools are rate limited.
