MCP
Integrate Midpage's legal database with AI assistants using the Model Context Protocol. Research opinions, trial filings, lawsuits, dockets, statutes, and regulations, and monitor cases and topics for new filings and decisions.
Versioning
These docs describe the current MCP contract, v5.
Midpage releases a new major version when an existing tool's inputs or outputs change or a tool is removed. A version can gain new tools and new optional fields. If your integration depends on fixed tool schemas or output fields, use a pinned version endpoint and ignore tools and fields it does not use.
Base URL
https://app.midpage.ai/mcp/v5
Use this pinned URL for the v5 contract documented here.
The unversioned URL follows the latest major version and may introduce breaking changes. It serves v5 today:
https://app.midpage.ai/mcp
Authentication
Midpage supports two authentication models for MCP integrations.
| Option | Best for | How it works |
|---|---|---|
| API key | Server-side integrations, internal tools, and quick prototypes | Your integration sends one shared API key with each request. |
| OAuth | End-user apps, partner integrations, and multi-tenant products | Each user signs in with their own Midpage account and your client sends that user's access token. |
1. API Key
Use API key auth for non-interactive access, server-side jobs, or any integration that should run under one shared credential.
Generate an API key in the Developer Portal. If you need help, contact support@midpage.ai. Then send it in your MCP request headers as:
Authorization: Bearer <api_key>
2. OAuth
Use OAuth when your client should connect each user to their own Midpage account. Each user will need a Midpage account before they can sign in and authorize your client.
For most customers, the easiest setup is:
- Point your client at
https://app.midpage.ai/mcp/v5 - Let your MCP or OAuth library follow Midpage's discovery metadata and register a client automatically if it supports dynamic client registration
- Let the user sign in and approve access when prompted
Most MCP clients should not need pre-provisioned OAuth app credentials. Midpage's Clerk auth server supports dynamic client registration, so compatible clients can create their own credentials automatically. If your client cannot do that and needs a pre-provisioned OAuth app, contact Midpage.
Need the manual OAuth settings?
If your client supports discovery, start there. In most cases, pointing it at https://app.midpage.ai/mcp/v5 is enough. Midpage publishes protected-resource metadata on app.midpage.ai, and that metadata points clients to the Clerk authorization server. The settings below are only for cases where you need to configure the OAuth flow yourself.
Discovery
- MCP protected-resource metadata:
https://app.midpage.ai/.well-known/oauth-protected-resource/mcp - Clerk OAuth metadata:
https://clerk.midpage.ai/.well-known/oauth-authorization-server - Clerk OIDC discovery:
https://clerk.midpage.ai/.well-known/openid-configuration
Some older MCP clients expect OAuth authorization-server metadata on the same origin as the MCP server. Midpage currently publishes protected-resource metadata on app.midpage.ai and authorization-server metadata on the Clerk domain above, so older clients may need manual configuration using those Clerk URLs.
The current metadata advertises:
- authorization server:
https://clerk.midpage.ai - dynamic client registration endpoint:
https://clerk.midpage.ai/oauth/register - PKCE challenge method:
S256 - resource scopes:
profile,email
Which OAuth flow to use
- Public clients such as desktop apps, mobile apps, browser apps, and local CLIs should use Authorization Code + PKCE (
S256). Do not embed aclient_secretin those clients. If you need this mode, Midpage can provision your Clerk OAuth app as a public client. - Confidential server-side apps can use Authorization Code and exchange the code on their backend with either
client_secret_basicorclient_secret_post. - If you need refresh tokens for long-lived sessions, request
offline_access.
Minimal flow
- Fetch the protected-resource metadata document and read
authorization_servers. - Fetch the Clerk authorization-server or OIDC metadata document and use the published endpoints.
- Start an authorization request with
response_type=code,client_id,redirect_uri,state, and theresourcevalue from the protected-resource metadata document. - Request
profile emailfor MCP access. Addopenidif your client expects an ID token. Addoffline_accessif you need a refresh token. - Public clients must also send
code_challengeandcode_challenge_method=S256. - Exchange the authorization
codefor tokens at the Clerk token endpoint, again including the sameresourcevalue. - Call the MCP server with
Authorization: Bearer <access_token>.
Example authorize request for a public client
GET https://clerk.midpage.ai/oauth/authorize?
response_type=code&
client_id=YOUR_CLIENT_ID&
redirect_uri=https://your-app.com/oauth/callback&
resource=https%3A%2F%2Fapp.midpage.ai&
scope=profile%20email%20openid%20offline_access&
state=RANDOM_VALUE&
code_challenge=BASE64URL_SHA256(code_verifier)&
code_challenge_method=S256
Example token exchange for a public client
curl -X POST "https://clerk.midpage.ai/oauth/token" \
-H "Content-Type: application/x-www-form-urlencoded" \
-d "grant_type=authorization_code" \
-d "client_id=YOUR_CLIENT_ID" \
-d "code=AUTHORIZATION_CODE" \
-d "redirect_uri=https://your-app.com/oauth/callback" \
-d "resource=https://app.midpage.ai" \
-d "code_verifier=YOUR_CODE_VERIFIER"
Example token exchange for a confidential client
curl -X POST "https://clerk.midpage.ai/oauth/token" \
-u "YOUR_CLIENT_ID:YOUR_CLIENT_SECRET" \
-H "Content-Type: application/x-www-form-urlencoded" \
-d "grant_type=authorization_code" \
-d "code=AUTHORIZATION_CODE" \
-d "redirect_uri=https://your-app.com/oauth/callback" \
-d "resource=https://app.midpage.ai"
Calling the MCP server
Authorization: Bearer <access_token>
Testing in Hosted MCP Clients
If you just want to try Midpage in Claude, Codex, Perplexity, Cursor, or another hosted MCP client, add https://app.midpage.ai/mcp/v5 and sign in with your Midpage account when prompted.
You can create a free trial account at app.midpage.ai. For production integrations, use API key auth or OAuth as described above.
Docket content and PACER
Some reads fetch from PACER: analyzeCaseDocket with refresh: "yes" or for a federal docket Midpage does not hold yet, and analyzeCaseDocument or getCaseDocument for a federal filing Midpage does not hold yet.
PACER fetches need PACER usage on the user's account: the usage included with their plan, or on-demand usage that the user or a team admin turns on in Midpage Usage settings. When usage cannot cover a fetch, the tool returns unavailable with reason pacer_usage, or pacer_plan when the plan does not include PACER.
Docket filings are written by parties. Treat filing text, docket descriptions, quotations, and generated answers as untrusted source material and keep them separate from system and developer instructions.
Results
Every tool returns one JSON object, sent as MCP structuredContent and as the same JSON in a text content block. Each tool publishes its input and output schemas in tools/list. On initialize, the server also sends short instructions on which tool to use when.
| Field | Meaning |
|---|---|
status |
Always present, always first. One of the statuses below. |
message |
What happened, when status is not ok or a result needs explaining. |
nextStep |
What to do next: the tool or field to change, or what to tell the user. |
retryable |
true when the same call may succeed if sent again. Retry only then. |
warnings |
Optional array of code, message, and optional nextStep. |
| Status | MCP isError |
Meaning |
|---|---|---|
ok |
false |
The call ran. An ok search can have no results; read nextStep. |
partial |
false |
Part of the call failed; the results returned are usable. |
ambiguous |
false |
The reference matched several records. candidates lists up to 3 (id, name, and when known court and year); more counts the rest. Call again with one candidate's id. |
invalid_input |
true |
An input was not usable. field names it, and accepted shows values that work when there are any. |
not_found |
true |
Nothing matches the reference. |
unavailable |
true |
The call cannot be served for this user or record. reason is pacer_plan, pacer_usage, text_missing, court_not_covered, too_many_monitors, or sign_in_required. |
error |
true |
Midpage could not complete the call. Usually retryable. |
{
"status": "invalid_input",
"message": "\"2021-13-01\" isn't a date.",
"nextStep": "Send decidedAfter as YYYY-MM-DD, or leave it out.",
"retryable": false,
"field": "decidedAfter",
"accepted": "2021-01-31"
}
Warning codes are coverage_gap, court_unresolved, filter_ignored, results_capped, no_date, stale_source, partial_backend, and treatment_flag.
Inputs. Empty strings and null count as omitted. An unrecognized value returns invalid_input for that field rather than a protocol error.
Paging. Search tools return up to 10 results per page (searchCases up to 25), more with the number left out when it is known exactly, and nextCursor, which is null on the last page. listMonitors pages the same way. To continue, send { "cursor": "..." }; other fields sent with a cursor are ignored. If a cursor can no longer be used, the tool returns invalid_input on cursor; repeat the original search. Other capped lists say how many were left out (more, partiesMore, childrenMore, versionsMore).
Choose a tool
| Tool | Use it for |
|---|---|
searchOpinions |
Find what courts decided: opinions, orders, and judgments, including state trial court rulings. |
searchFilings |
Search the text of what parties filed in state trial courts. |
searchCases |
Find lawsuits by party, attorney, firm, judge, docket number, court, case type, or filing date. |
analyzeCaseDocument |
Answer a question about one opinion or filing, with verified quotations and pin cites. |
getCaseDocument |
Read one opinion or filing in full, with a link to its original file. |
analyzeCaseDocket |
Answer a question about one case from its docket and locate its filings. |
searchLaws |
Find statutes, regulations, constitutions, the federal rules of procedure and evidence, and agency guidance. |
analyzeLaw |
Answer a question about one provision with verified passages. |
createCaseDocketMonitor |
Email new filings on one case's docket. |
createSearchMonitor |
Run a search on a schedule and email what is new. |
listMonitors |
List the user's monitors and what each last found. |
updateCaseDocketMonitor, updateSearchMonitor |
Change a monitor's recipients, schedule, or instruction, or pause and resume it. |
deleteMonitor |
Delete a monitor and stop its emails. |
Search results and docket descriptions are for discovery. Read a document with analyzeCaseDocument before citing it. Treatment appears only when the citator found Negative or Caution history; its absence does not establish that an authority is good law.
searchOpinions
Find what US courts decided: opinions, orders, and judgments, including state trial court rulings when a state's trial courts are in scope. Send up to four queries for different angles of one question, each with its own filters. Results are merged into one ranking, and a hit several queries find ranks higher.
Input
| Field | Type | Meaning |
|---|---|---|
queries |
object[] | 1 to 4 queries, each with the fields below. |
queries[].query |
string, required | Legal issue, terms, case name, citation, or quoted text. |
queries[].courts |
string[] | Bluebook abbreviations or full names, OR-ed (9th Cir., S.D.N.Y., Tex. App.). |
queries[].states |
string[] | Full state names: their state courts, the federal courts sitting there, and their circuit. |
queries[].courtSystem |
string | all (default), state, or federal. Narrows states, or the whole search. |
queries[].decidedAfter, queries[].decidedBefore |
string | Inclusive decision dates, YYYY-MM-DD. |
queries[].publishStatus |
string | any (default), published, or unpublished. |
queries[].citing |
string | Only opinions that cite this document. See Citing. |
cursor |
string | nextCursor from the last page, sent alone. |
Values in one list are OR-ed; different filters are AND-ed.
{
"queries": [
{ "query": "PSLRA safe harbor statements of current fact", "courts": ["9th Cir."] },
{ "query": "forward-looking statement mixed with present facts", "states": ["California"] }
]
}
Output
| Field | Meaning |
|---|---|
results |
Up to 10 opinions: documentId, caseId, caseName, citation (the full citation, like Bell Atl. Corp. v. Twombly, 550 U.S. 544 (2007), or No. {docket} when Midpage has none), court, dateDecided, url, a snippet of at most 440 characters, and treatment (Negative or Caution) when the citator flagged it. |
more, nextCursor |
See Paging. Up to 100 matches can be paged through. |
Citing
Set citing on a query to a reporter citation, a documentId, or a Midpage document URL to search only the opinions that cite that document. Combine it with the query's other filters, for example to find Ninth Circuit opinions applying a case to a particular issue, or to read the decisions behind a treatment flag.
- A query with
citingdoes not search state trial court rulings. - To list every opinion citing a document, send
"query": "*"withciting. - A target Midpage cannot match returns
invalid_inputonqueries[i].citing. Copies of one decision printed at the same citation are searched together. A citation printed for several different cases returnsinvalid_inputwith up to 3 candidates inaccepted, each itsdocumentId, case name, court and year. - For a widely cited opinion, only its most recent citing opinions are searched, with a
results_cappedwarning. AdddecidedAfterordecidedBeforeto search the rest.
{
"queries": [
{
"query": "plausibility standard applied to antitrust conspiracy",
"citing": "550 U.S. 544",
"courts": ["9th Cir."]
}
]
}
searchFilings
Search the text of what parties filed in state trial courts: complaints, petitions, answers, motions, briefs, and exhibits. Coverage is deepest in New York, Texas, Florida, and California. Federal filing text is not searchable: find a federal case with searchCases, open it with analyzeCaseDocket, and read an entry with analyzeCaseDocument.
Input
| Field | Type | Meaning |
|---|---|---|
queries |
object[] | 1 to 4 queries, each with the fields below. |
queries[].query |
string, required | Issue, key terms, or a quoted phrase. |
queries[].state |
string | One full state name. Required unless courts names that state's trial courts. |
queries[].courts |
string[] | Trial courts in that state, OR-ed. Defaults to statewide. |
queries[].documentTypes |
string[] | complaint, petition, answer, motion, brief, exhibit, other, OR-ed. Defaults to every type. |
queries[].filedAfter, queries[].filedBefore |
string | Inclusive document filing dates, YYYY-MM-DD. |
cursor |
string | nextCursor from the last page, sent alone. |
{
"queries": [
{
"query": "motion to compel arbitration delegation clause",
"state": "Texas",
"documentTypes": ["motion", "brief"]
}
]
}
Output
| Field | Meaning |
|---|---|
results |
Up to 10 filings: documentId, documentType (complaint, answer, motion, brief, order, judgment, exhibit, or other), title, caseId, caseName, court, docketNumber, dateFiled, url, and snippet. |
more, nextCursor |
See Paging. |
A filing states a party's position, not a ruling.
searchCases
Find lawsuits by their docket metadata. There is no subject search: narrow by names, courts, and dates. Send at least one filter. Different fields are AND-ed; values in a list are OR-ed.
Input
| Field | Type | Meaning |
|---|---|---|
caseName |
string | Words in the caption. |
parties |
string[] | Up to 20 names, each matched as a phrase in party names or the caption. Tesla finds Tesla, Inc.. |
attorneys, firms, judges |
string[] | Names. Mostly state court data. Attorney and judge names match in either order: Wesley Poole finds POOLE, WESLEY R. |
docketNumber |
string | Any common form. Case, punctuation, spacing, and a leading No. are ignored; in a federal number, so are leading zeros and the judge's initials. |
courts |
string[] | Bluebook abbreviations or full names, OR-ed. |
states |
string[] | Full state names: courts sitting in those states, state and federal. |
courtSystem |
string | all (default), state, or federal. |
courtLevel |
string | all (default), trial, or appellate. Courts in U.S. territories have no recorded level, so trial and appellate leave them out. |
caseTypes |
string[] | civil, criminal, bankruptcy, vaccine. Recorded for federal cases only, so it leaves out state cases, with a coverage_gap warning. For appeals, use courtLevel. |
filedAfter, filedBefore |
string | Inclusive case filing dates, YYYY-MM-DD. For older appellate cases, the decision date. |
sort |
string | newest (default) or relevance. |
cursor |
string | nextCursor from the last page. |
{
"parties": ["Tesla"],
"states": ["Texas"],
"filedAfter": "2026-11-01"
}
Output
| Field | Meaning |
|---|---|
results |
Up to 25 cases: caseId, caseName, court, docketNumber, url, and when known dateFiled, caseType, status, and parties (the first 3, with partiesMore counting the rest). |
total |
All matches, counted up to 10,000. totalIsLowerBound is true when there are more. |
byCourt |
The 5 courts with the most matches, with counts. |
nextCursor |
See Paging. |
Federal cases list parties only in the caption. Party roles are not recorded. Cases without a filing date drop out of date filters, with a no_date warning. Pass caseId to analyzeCaseDocket for the docket, or to createCaseDocketMonitor to follow the case.
analyzeCaseDocument
Read one opinion or filing and answer a question about it with verified quotations and pin cites. To read the full text, use getCaseDocument.
Input
| Field | Type | Meaning |
|---|---|---|
ref |
string, required | A documentId from any tool, a reporter citation (550 U.S. 544), or a Midpage document URL. For a docket entry, a caseId with entryNumber. WL and LEXIS numbers are not supported. |
entryNumber |
string | Docket entry number, with a caseId as ref. |
attachmentNumber |
string | Attachment number within that entry. Omit for the main document. |
question |
string, required | Everything you need from this document. Several parts are fine. |
{
"ref": "556 U.S. 662",
"question": "Are threadbare recitals of a cause of action entitled to an assumption of truth?"
}
Output
| Field | Meaning |
|---|---|
document |
documentId, caseId, caseName, court, date, docketNumber, citation, and reader url. For an opinion, publishStatus and, when they apply, perCuriam, enBanc, and precedentialNote. For a filing, documentType, entryNumber, attachmentNumber, and description. |
answer |
A direct answer, at most 6,000 characters. |
supportedPropositions |
Up to 20, most relevant first: proposition (with its conditions), quote (verbatim, at most 1,600 characters), citation (for an opinion, its citation with the pin cite when Midpage has the page, like Bell Atl. Corp. v. Twombly, 550 U.S. 544, 556 (2007)), deeplinkUrl (a link to the passage, when Midpage has one), and opinionSection when not the majority opinion. more counts the rest; ask a narrower question for them. |
doesNotAddress |
Parts of the question the document does not answer. Do not cite it for those. |
treatment |
Only when the citator found Negative or Caution history: signal, counts (negative, caution), up to 3 citing opinions (documentId, citation), and more. A treatment_flag warning comes with it. |
Cite a proposition or quote with its citation, exactly as returned. A federal entry Midpage does not hold may be fetched from PACER. See Docket content and PACER.
getCaseDocument
Read one opinion or filing in full: verbatim text in pages of up to 60,000 characters, and a link to the original file. A federal entry Midpage does not hold may be fetched from PACER. See Docket content and PACER.
Input
| Field | Type | Meaning |
|---|---|---|
ref, entryNumber, attachmentNumber |
string | As in analyzeCaseDocument. |
offset |
integer | nextOffset from the last page. Defaults to 0. |
includeText |
string | yes (default) or no. no returns only the citation and file link. |
{
"ref": "550 U.S. 544"
}
Output
| Field | Meaning |
|---|---|
document |
documentId, caseId, caseName, court, date, citation, documentType, and reader url. |
text |
Up to 60,000 characters. |
nextOffset |
Send as offset for the next page. Absent at the end. |
paginationCitation |
The reporter whose pages the [*N] markers count. |
downloadUrl |
The original file, when Midpage holds one. |
Opinion text is Markdown: ## opinion parts, > quoted material, [FNn] at a footnote's reference and again where its note begins, and [*N] where page N of paginationCitation begins. [*~N] is an approximate page break; check the file before pin-citing beside it. [*?], or no paginationCitation, means pages are unknown. Filing text is plain text.
analyzeCaseDocket
Answer a question about one known case from its docket: parties, lawyers, judge, case type, what was filed and when, how motions were decided, and status. It also locates filings for analyzeCaseDocument.
Input
| Field | Type | Meaning |
|---|---|---|
ref |
string | A caseId or a Midpage docket URL. Send this, or court and docketNumber. |
court |
string | Bluebook abbreviation or full name, with docketNumber. |
docketNumber |
string | Docket number as written, with court. |
question |
string, required | What to find out about the case. |
refresh |
string | no (default) uses stored data. yes fetches a fresh docket from PACER; if PACER refuses, the stored docket comes back as partial with the reason. |
{
"court": "N.D. Ill.",
"docketNumber": "1:16-cv-11057",
"question": "What was filed or entered during the past month?",
"refresh": "yes"
}
Output
| Field | Meaning |
|---|---|
case |
caseId, caseName, court, docketNumber, dateFiled, status (open or closed, as the docket records it; absent when it doesn't say) with dateTerminated once the case ends, and the docket url. |
answer |
Grounded in the docket, at most 6,000 characters. |
entries |
Up to 40 entries that bear on the question: entryNumber, attachmentNumber, dateFiled, description, documentId when Midpage holds the filing, and stored: false when reading it requires a PACER fetch. more counts the rest. |
asOf, fetchedAt |
The newest entry date and when the docket was retrieved. Neither proves the docket is current. |
doesNotAddress |
Parts of the question the docket does not answer. |
downloadUrl |
A link to the docket report. |
To read an entry, pass its documentId as ref to analyzeCaseDocument, or the caseId as ref with entryNumber and attachmentNumber.
searchLaws
Find US federal and state statutes, regulations, constitutions, the federal rules of procedure and evidence, and agency guidance by concept, text, citation, or popular name. Midpage has no state court rules. Pass a result's lawId to analyzeLaw before citing it.
Input
| Field | Type | Meaning |
|---|---|---|
query |
string | Concept, terms, citation, popular name, or quoted text. Required unless cursor. |
states |
string[] | Issuing jurisdictions: full state names, or Federal. |
collectionTypes |
string[] | statute, regulation, constitution, court_rule, guidance, OR-ed. guidance includes executive orders and notices. The federal rules of procedure and evidence are under statute. |
within |
string | One code or title by name, like Cal. Lab. Code or 29 C.F.R.. |
includeHistorical |
string | no (default) or yes to add historical and future-effective versions. |
effectiveDateFrom, effectiveDateTo |
string | Effective date range, inclusive, YYYY-MM-DD. Statutes and regulations are dated by effective date; laws without one are left out. |
publicationDateFrom, publicationDateTo |
string | Publication date range, inclusive, YYYY-MM-DD. Agency guidance and register documents are dated by publication date; documents without one are left out. |
cursor |
string | nextCursor from the last page. |
{
"query": "itemized wage statement requirements",
"states": ["California"],
"collectionTypes": ["statute"]
}
Output
| Field | Meaning |
|---|---|
results |
Up to 10 provisions: lawId, citation, title, jurisdiction, collectionType (statute, regulation, constitution, court_rule, guidance, executive_order, or notice), isCurrent, effectiveDate, publicationDate when known, url (that exact version), and snippet. |
byJurisdiction |
Estimated matches in the 10 jurisdictions with the most; byJurisdictionMore counts the others. |
more, nextCursor |
See Paging. |
analyzeLaw
Answer a question about one provision with verified passages from its text.
Input
| Field | Type | Meaning |
|---|---|---|
ref |
string, required | A lawId (preferred), or a canonical citation, Federal Register citation, or Midpage law URL. |
question |
string, required | A specific question about this provision. |
includeHistorical |
string | no (default) or yes to let a citation match non-current versions. A lawId always returns that exact version. |
{
"ref": "21 C.F.R. ยง 820.30",
"question": "What design control requirements apply to medical device manufacturers?"
}
Output
| Field | Meaning |
|---|---|
law |
lawId, citation, title, jurisdiction, isCurrent, effectiveDate, parentId, registerCitation when it has one, url, and sourceUrl (the official publisher's page). |
answer |
A compact answer. |
passages |
Up to 8 verified passages: point, verbatim quote, and url. more counts the rest. |
doesNotAddress |
Parts of the question the provision does not answer. |
children |
For a title or chapter without its own text: up to 250 children (lawId, citation, title); childrenMore counts the rest. |
versions |
Up to 250 versions (lawId, effectiveDate, isCurrent); versionsMore counts the rest. |
Monitors
A monitor emails its recipients when something new arrives: a new filing on a case's docket, or new opinions and filings that match a standing instruction. Each monitor belongs to the signed-in user, so these tools need OAuth or a hosted client sign-in. With an API key they return unavailable with reason sign_in_required.
recipients is a list of up to 10 email addresses, with "me" for the user. Omit it to email only the user. On update it replaces the whole list.
| Tool | Input | Output |
|---|---|---|
createCaseDocketMonitor |
caseId (required), recipients |
created and monitor. A case already monitored returns that monitor with created: false. |
createSearchMonitor |
title, prompt (the standing instruction), cron (five fields, at most hourly), all required; timezone (IANA, default America/New_York), recipients |
monitor. A user can have 10 search monitors. |
listMonitors |
cursor |
monitors, newest first, and nextCursor. |
updateCaseDocketMonitor |
monitorId (required), recipients, active (false pauses, true resumes) |
monitor. Omitted fields stay as they are. |
updateSearchMonitor |
monitorId (required), recipients, active, title, prompt, cron, timezone |
monitor. Changes apply from the next run. |
deleteMonitor |
monitorId (required) |
deleted: true. Deleting a search monitor also deletes its thread in Midpage. |
{
"title": "Weekly MTD watch",
"prompt": "New Ninth Circuit opinions deciding motions to dismiss under Rule 12(b)(6) in securities fraud cases.",
"cron": "0 8 * * 1",
"timezone": "America/Los_Angeles"
}
A case docket monitor in a court with no free feed of new filings is checked on PACER twice a day, and each check counts toward the user's PACER usage. The create result says so in a warning, and listMonitors warns when checks stopped because usage could not cover them.
Monitor object
| Field | Meaning |
|---|---|
monitorId, type |
type is caseDocket or search. |
active |
false when paused. A paused monitor sends nothing. |
recipients, createdAt |
Who gets the emails, and when the monitor was created. |
case |
Case docket monitors: caseId, caseName, court, docketNumber, and the docket url. |
paid |
Case docket monitors: true when the docket is checked on PACER. |
checksStoppedAt |
Case docket monitors: when PACER checks stopped. They resume once usage can cover them. |
title, prompt, cron, timezone, nextRunAt |
Search monitors: what it looks for and when it runs next. |
url |
Search monitors: its thread in Midpage. |
lastRun, lastReport |
Search monitors, once run: lastRun has at and status (running, found, nothing_new, or failed); lastReport has at, headline, and the Markdown report. |
Failures use the shared statuses: invalid_input for a bad recipients, cron, or id, not_found for an unknown monitor or case, and unavailable with reason court_not_covered, pacer_plan, too_many_monitors, or sign_in_required.
Migrate from v4
| V4 | V5 |
|---|---|
search |
searchOpinions for court rulings, searchFilings for state trial court party filings. Both keep queries[] with per-query filters; the cursor moves to a top-level cursor. |
jurisdiction, startDate, endDate, documentTypes |
states plus courtSystem; decidedAfter/decidedBefore or filedAfter/filedBefore; searchFilings takes seven filing types. |
| No lawsuit search | searchCases. |
analyzeCaseDocument with documentId, citation, or entry |
One ref, plus entryNumber and attachmentNumber. The case and document fields are merged into one document; treatment appears only when flagged. |
getCaseDocument, 60,000-character pages |
ref, which also takes a docket entry, the same 60,000-character pages, includeText: "yes" or "no". |
analyzeCaseDocket with caseId or docket, forceFetch |
ref, or court and docketNumber; refresh: "no" or "yes". relevantEntries becomes entries. |
searchLaws with collectionTypes, collections, pathPrefix, currentOnly, page |
collectionTypes (adds court_rule; as a filter, guidance includes executive orders and notices, and results name them executive_order and notice), within, includeHistorical, cursor. The date filters keep their v4 names. id becomes lawId; stateCounts becomes byJurisdiction. A v4-only field is ignored with a filter_ignored warning. |
analyzeLaw with citation, registerCitation, path, or id |
One ref. |
| No monitor tools | createCaseDocketMonitor, createSearchMonitor, listMonitors, updateCaseDocketMonitor, updateSearchMonitor, and deleteMonitor. |
Per-tool status and error, JSON text only |
The Results contract, with structuredContent and output schemas. |
V5 is not a drop-in replacement: update tool names, references, and response parsing together. Keep integrations pinned to /mcp/v4 until they are ready to migrate.
Older versions
Older contracts keep their pinned endpoints (https://app.midpage.ai/mcp/v4, and likewise v3, v2, v1) and their reference pages: v4, v3, v2, v1. The MCP changelog has the version history.