proofread-mcp
proofread-mcp is an MCP server that checks case citations in drafts against proofread.law's registers of ~10M US court opinions and 1.1M Swiss decisions, and suggests leading cases.
Check citations in prose (
check_citations): paste text (up to 2M chars); returns coverage statement, per-tier counts, flagged rows, citation count and a report id. Optionaldeep: truealso judges whether each cited opinion supports the sentence it is cited for.Check citations in a file (
check_document): same result for a local.pdf,.docx,.txtor.mdfile up to 10 MB (.docxneeds a paid plan).Look up one citation (
resolve_citation): answers whether a case sits at a citation (found, ambiguous, not in register, cannot verify, known citation, or no citation recognised), plus that volume's coverage.Look up a list of citations (
resolve_citations): up to 500 citation strings in one call, one line per citation in input order.See what's covered (
coverage): the coverage statement and storage notice;jurisdiction: "ch"gives the Swiss register, courts held and index share.Render a report (
render_report): markdown diligence report from a report id or full report JSON, with every row.Suggest Swiss leading cases (
suggest_cases): from a Swiss federal statute article (e.g.Art. 41 OR) or a paragraph citing one, get a ranked list of leading BGE cases with field of law, quoted passage, practice changes and links; filters for domain, language andk.Suggest US cases (beta) (
suggest_cases): from a sentence stating a rule of law, get up to 3 cases whose text states it, with the matched passage, how the case binds a named court, and opinion links.Save and manage briefs (opt-in):
save_brief,list_briefs,get_brief,update_brief(edit-and-recheck loop, keeps last 20 versions) anddelete_brief; stored encrypted in the user's proofread.law account and requires an API key.Account and billing:
sign_upcreates an account and API key (shown once, adopted for the session);billing_linkreturns a Stripe Checkout link for payg, solo or firm plans.Limits to know: cannot resolve Westlaw/Lexis identifiers, check statutes, regulations or secondary sources, or say whether a case is still good law; a red row means "check this," never "this case does not exist."
Generates Stripe Checkout links for upgrading a proofread.law account to a paid plan (pay as you go, solo, or firm).
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@proofread-mcp Check the case citations in this brief before we file."
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
proofread-mcp
An MCP server for proofread.law. It lets Claude Desktop, Claude Code, Cursor and the OpenAI Agents SDK check case citations before a draft is filed: about 10 million US court opinions, and 1.1 million Swiss decisions (BGE/ATF/DTF, Federal Supreme Court dockets, the federal courts and all 26 cantons). The jurisdiction is detected from the draft, and a Swiss report comes back in the draft's language: German, French or Italian.
proofread.law checks each citation against an open register of about 10 million court opinions (CourtListener bulk data). Every result says what was checked, what was found, and what the register cannot see.
What it does
Fourteen tools:
Tool | Input | What comes back |
| text, | The coverage statement, counts per tier, one line per row that needs a human (red first, then orange, then deep-check rows to review), the number of citations found, a report id |
| path to a | The same, for a file on disk |
| one citation string | The register's answer for that citation: found (case, court, date, parallel citations, link), ambiguous (candidates), not in the register, cannot verify, known citation, or no citation recognised; the coverage of that volume; the coverage statement |
| a list of up to 500 citation strings | Counts by status, one line per citation in input order, the coverage statement |
|
| The coverage statement and the storage notice; |
| a report id from a previous check, or the full report JSON | A markdown diligence report with every row |
| a Swiss federal statute article or a paragraph that cites one, or a sentence of US law; | Swiss: the leading Federal Supreme Court cases (BGE) cited with the article, to read, in one ranked list, each labelled with its field of law and given with the quoted passage, any later change of practice, the decision's link and a link to check the citation; in the query's language. US (beta): up to 3 cases whose own text states the sentence, each with the passage the model matched, the paragraphs around it, how it binds the court you name, the opinion's link and a link to check the citation |
| text, | Saves the brief to the user's account (opt-in, stored encrypted) and checks it: the brief id, then the same compact result as |
| none | The saved briefs: id, title, when saved and last checked, counts per tier, number of versions |
| id, | The saved text, the latest report and the versions kept; with |
| id, | Saves the edited text as a new version and re-checks it: the flags resolved since the previous version, the new flags (each with its headline), the unchanged count, then every row that still needs attention. |
| id | Permanently deletes the brief and its versions |
| the account owner's email, a name for the agent | A proofread.law account and an API key (shown once); the server uses it for the rest of the session |
|
| A Stripe Checkout link for the account owner; needs an API key |
The compact result of a check is capped at about 12,000 characters; when a long brief has more flagged rows than fit, the text says how many were left out and render_report has them all.
Saved briefs (opt-in)
Nothing is saved unless save_brief or update_brief is called; check_citations and check_document never save anything. A saved brief is stored encrypted in the user's proofread.law account until delete_brief removes it, so the brief tools need an API key (any PROOFREAD_API_KEY, or one from sign_up); without one they answer with a message and send nothing. A server with storage switched off answers storage_off, and the tools say so.
The loop for fixing flagged citations: save_brief once, fix a flagged row in the text (the citation, the case name, the quotation, or take the citation out), call update_brief with the id and the whole edited text, and read what changed: which flags were resolved (flagged before, not flagged now), which are new, and the rows that still need attention. The last 20 versions are kept; get_brief lists them and reads an earlier one. Ids are integers in the API and strings in the tools.
check_citations and check_document read prose: they compare the case name and any quotation with the register. resolve_citation and resolve_citations look the citation string up in the register (the /v1/resolve API) and tell you which case sits there; they do not compare it with the name you have.
Tiers, in the words the tools use:
Tier | Word | Meaning |
red | check this | The register holds something concrete that disagrees: a different case at that citation, a volume or page that does not match, quoted words not in the opinion |
orange | cannot verify | Nothing to check against: a Westlaw or Lexis identifier, a volume newer than the register, a reporter the register holds only in part. Not evidence either way |
green | found | The citation resolves to a case whose caption matches |
white | deep check | With |
What it cannot do: resolve Westlaw (WL) or Lexis identifiers, check statutes, regulations or secondary sources, or say whether a case is still good law. Those limits are stated in every tool description and in the coverage statement that comes with every result.
Case suggestions (Switzerland)
suggest_cases takes a Swiss federal statute article (Art. 41 OR, art. 41 CO, Art. 8 ZGB, art. 9 Cst.; lowercase and dotless forms such as art 41 or are read too) or a paragraph that cites articles, and lists the leading Federal Supreme Court cases (BGE) the court cites with it. Federal acts only; cantonal law is not covered. There is no free-text search: a query without an article answers with a message that no article was recognised. The answer is one ranked list in the measured order, each case labelled with its field of law; the domain filter keeps the same order within one field, so a filtered row keeps its overall rank (Rang 7, Rang 9, ...) and a filter note says which field is shown. Each row has the citation, date, field, the decision's language, its rank, the quoted passage (regeste or consideration), any later change of practice (a changed precedent is listed with its flag, never dropped), the decision at the court's site and a link that opens the check on proofread.law. Every string comes back in the query's language (German, French or Italian; lang: "en" for English), and the tool prints it as it is. A suggestion is a case to read: it has not been checked against your sentence, and a row without a flag is not evidence that its practice still holds. To check a citation, use check_citations.
Arguments: query (required, up to 20,000 characters; over 1,500 it is sent as POST), domain (all, civil, criminal, public or social; default all), lang (de, fr, it or en; default the query's language), k (number of rows, 1 to 50; default 10). With a large k, rows past about 20,000 characters of text keep their header line and any practice flag and leave out the passage and links, which stay in the structured result (with the counts per field). During the trial phase it needs a paid-plan or trial API key; otherwise the API answers plan_required and the tool says where to upgrade.
Example call, suggest_cases with {"query": "Art. 41 OR, Art. 97 OR", "k": 3} (dev instance, 2026-09-26; rows 2 and 3 cut):
Gelesen als: Art. 41 OR, Art. 97 OR
1. BGE 146 IV 76, 13.11.2019, Strafrecht, FR, Rang 1
Das Bundesgericht zitiert diesen Entscheid zusammen mit Art. 41 OR; der Entscheid selbst nennt den Artikel nicht. Aus der Regeste: «a) Art. 110 Abs. 1 StGB; Art. 118, 121 Abs. 1 und 382 Abs. 1 StPO; ...»
Entscheid öffnen: https://search.bger.ch/ext/eurospider/live/de/php/clir/http/index.php?highlight_docid=atf%3A%2F%2F146-IV-76%3Ade&lang=de&type=show_document
Zitat prüfen: https://proofread.law/?cite=Art.%2041%20OR%3B%20BGE%20146%20IV%2076#check
...
Die Vorschläge sind publizierte Leitentscheide (BGE). Sie sind danach gereiht, wie oft das Bundesgericht sie zusammen mit dem Artikel zitiert, [...] Ein Vorschlag ist ein Entscheid zum Lesen: Ob er Ihre Aussage stützt, prüft diese Liste nicht, und ein Entscheid ohne Hinweis ist kein Beleg dafür, dass seine Praxis weiter gilt.With "domain": "civil" the list starts with the filter note and keeps the overall ranks:
Gelesen als: Art. 41 OR
Filter: Zivilrecht. Die Reihenfolge ist dieselbe wie in der ganzen Liste.
1. BGE 132 III 122, 13.09.2005, Zivilrecht, FR, Rang 7
Regeste: «Rechtmässigkeit von im Arbeitskampf eingesetzten Mitteln (Art. 28 BV; Art. 41 und 357a OR). ...»
...A filter that leaves no rows answers with a message saying so. A later change of practice appears right under the row it concerns, for example Praxisänderung durch BGE 145 III 1, möglicherweise nur teilweise: <link> followed by the regeste sentence that marks it.
Case suggestions (United States, beta)
suggest_cases also takes a sentence from a US draft that states a rule of law (one sentence, or a short paragraph up to 1,500 characters) and lists up to 3 cases whose own text states it. A model reads the opinion of the court in each of the top 10 candidates, and a case is shown only when the model finds a passage in it that states the sentence. Each row has the reference, the court, the year, how the case stands to the court you name (binding here, same circuit, persuasive), any later history the citator found (reversed in part, superseded by statute), the passage the model matched with the paragraphs before and after it (each cut to about 600 characters), a link to the opinion on CourtListener and a link that opens the check on proofread.law. Cases that a later court overruled or reversed are left out, and a note names them.
This is a beta. Every answer starts with a line that says so, with the measured numbers: on the site's own index, of 94 cases shown for 50 sentences from filed briefs, 73 (78%) support the sentence and 4 do not, as read blind by two model readers, not practising lawyers (details). Read the passage before you cite the case. A suggestion is a case to read; to check a citation, use check_citations.
Arguments: court (optional, up to 200 characters) is the court the brief is filed in: a federal court of appeals (9th Cir., Ninth Circuit, ca9), a district court (N.D. Cal., S.D.N.Y., cand), or a state or territory name or postal code (California, CA) for the federal court there. Cases that bind that court come first; without a court, Supreme Court cases come first. State courts are not read yet; a court the API does not read comes back as an error that says what it reads. jurisdiction is auto by default: a Swiss statute article, or German, French or Italian text, is Swiss, and anything else is a US sentence; us or ch sets it. domain, lang and k apply to Swiss answers only, and a US answer is in English.
Input that states no rule of law (a question, a citation or case name, a heading, a statement about the record or a party's argument, or more than 1,500 characters) gets no list. The answer says what to paste instead and gives an example sentence, and it costs nothing. A US answer that ran the model counts as one deep-checked citation, not a resolve. When the day's budget for model calls is spent, US suggestions are back the next day (UTC).
Example call, suggest_cases with a sentence on the plausibility pleading standard and "court": "N.D. Cal." (dev instance, 2026-09-26; the passage and the paragraphs around it shortened, rows 2 and 3 cut):
Beta: Each case is shown with the passage we matched; read it before you cite the case. Measured on the site's own index: 78% of shown cases support the sentence, and 4 of 94 did not (details on https://proofread.law/measurements#suggestions).
Ordered for N.D. Cal. (Ninth Circuit): binding cases first.
1. Bell Atlantic Corp. v. Twombly, 550 U.S. 544 (2007), Supreme Court, 2007, Binding here (Supreme Court)
The passage we matched: "Here, in contrast, we do not require heightened fact pleading of specifics, but only enough facts to state a claim to relief that is plausible on its face. Because the plaintiffs here have not nudged ..."
Before: "Plaintiffs say that our analysis runs counter to Swierkiewicz, 534 U. S., at 508 , which held that “a complaint in an employment discrimination lawsuit [need] n..."
After: "The judgment of the Court of Appeals for the Second Circuit is reversed, and the case is remanded for further proceedings consistent with this opinion...."
Read the opinion: https://www.courtlistener.com/opinion/145730/bell-atlantic-corp-v-twombly/
Check this citation: https://proofread.law/?cite=Bell%20Atlantic%20Corp.%20v.%20Twombly%2C%20550%20U.S.%20544%20%282007%29#check
2. Ashcroft v. Iqbal, 556 U.S. 662 (2009), Supreme Court, 2009, Binding here (Supreme Court)
...
3. Greg Landers v. Quality Communications, Inc., 771 F.3d 638 (9th Cir. 2014), Court of Appeals for the Ninth Circuit, 2014, Binding here
...
How we build this list: we search the sentences in which federal appellate courts and the Supreme Court cite a case, and the cases' own text. [...] We leave out cases that a later court overruled or reversed.A question gets the beta line, then the API's message and an example:
This reads as a question. Paste the rule itself, as a sentence, and we will look for cases that state it.
For example: "A complaint must contain sufficient factual matter, accepted as true, to state a claim to relief that is plausible on its face."Related MCP server: CourtListener Citation Validation MCP Server
Install
Needs Node 20 or newer. No install step is required; npx fetches proofread-mcp from npm.
Claude Desktop (extension bundle)
Download proofread-mcp-<version>.mcpb from the latest GitHub release and open it with Claude Desktop (double-click, or Settings, Extensions, Install Extension). The bundle carries the server and its dependencies and uses the Node runtime that ships with Claude Desktop, so nothing else has to be installed. The only setting is the optional API key, stored by Claude Desktop as a sensitive value and passed to the server as PROOFREAD_API_KEY.
To build the bundle yourself: scripts/build-mcpb.sh (needs npx @anthropic-ai/mcpb) writes build/proofread-mcp-<version>.mcpb from manifest.json, icon.png and the npm package contents.
Claude Desktop (config file)
Edit claude_desktop_config.json (Settings, Developer, Edit Config):
{
"mcpServers": {
"proofread": {
"command": "npx",
"args": ["-y", "proofread-mcp"],
"env": {
"PROOFREAD_API_KEY": "pl_..."
}
}
}
}Leave out env to use the free tier.
Claude Code
claude mcp add proofread -- npx -y proofread-mcp
# with an API key:
claude mcp add proofread -e PROOFREAD_API_KEY=pl_... -- npx -y proofread-mcpCursor
Settings, MCP, Add new global MCP server, or write .cursor/mcp.json in the project:
{
"mcpServers": {
"proofread": {
"command": "npx",
"args": ["-y", "proofread-mcp"],
"env": { "PROOFREAD_API_KEY": "pl_..." }
}
}
}OpenAI Agents SDK (over HTTP)
Start the server with the HTTP transport:
PROOFREAD_API_KEY=pl_... npx proofread-mcp --http --port 3333
# MCP endpoint: http://127.0.0.1:3333/mcp health: http://127.0.0.1:3333/healthThen connect from the Agents SDK:
from agents import Agent, Runner
from agents.mcp import MCPServerStreamableHttp
# The SDK's default read timeout is 5 s. A check of a long brief takes up to 6 s and a deep check 1 to 2 s per citation,
# so give the session up to 15 minutes (the API's own deep-check limit).
async with MCPServerStreamableHttp(params={"url": "http://127.0.0.1:3333/mcp"}, client_session_timeout_seconds=900) as proofread:
agent = Agent(name="Drafting assistant", instructions="Check every case citation before you rely on it.", mcp_servers=[proofread])
result = await Runner.run(agent, "Check the citations in this paragraph: ...")import { Agent, run, MCPServerStreamableHttp } from "@openai/agents";
const proofread = new MCPServerStreamableHttp({ url: "http://127.0.0.1:3333/mcp", name: "proofread" });
await proofread.connect();
try {
const agent = new Agent({ name: "Drafting assistant", mcpServers: [proofread] });
const result = await run(agent, "Check the citations in this paragraph: ...");
} finally {
await proofread.close();
}The HTTP server binds to 127.0.0.1 by default and is single-tenant by design: it has no authentication of its own, report ids are shared across sessions, and a key from sign_up is adopted by the whole process. Sessions that stay idle for 30 minutes are closed. To expose it on a network use --host 0.0.0.0 and put it behind something that adds authentication.
Any MCP client
stdio: run proofread-mcp. Streamable HTTP: run proofread-mcp --http --port 3333 and point the client at /mcp.
Environment variables
Variable | Default | Meaning |
| unset | An API key ( |
|
| Base URL, for a self-hosted or test instance |
Accounts and keys
An agent can open an account itself: sign_up posts the owner's real email and a name to POST /agent/signup and gets a key back, shown once. The server uses that key for the rest of the session; put it in PROOFREAD_API_KEY to keep it. The owner receives one confirmation email. When a quota is used up (a tool answers "needs the ... plan" or "monthly allowance used"), billing_link returns a Stripe Checkout link for the owner; nothing is charged until they pay. Plans and prices: proofread.law/pricing. The onboarding text the API publishes for agents is at proofread.law/agent/onboarding.md.
Free tier
Per month, per IP address without a key or per account with a free-tier key:
Tools | Quota |
| 20 checks, of which 3 may be deep checks |
| each runs a default check and counts as one of those checks; a title alone is not counted. Each plan has a cap on saved briefs |
| 1,000 resolves (each citation in a list counts as one) |
| paid plans and trials only during the trial phase; a Swiss answer counts as one resolve (a query without an article costs nothing); a US answer that ran the model counts as one deep-checked citation (input that states no rule of law costs nothing) |
| free, not counted |
There is also a limit of 20 requests an hour per IP (more on paid plans). When a limit is reached the tool returns a plain message with the retry time or the upgrade link; nothing is thrown at the protocol level.
.docx upload and unlimited checks need a paid plan. See proofread.law/pricing.
Privacy Policy
The full policy is at proofread.law/privacy. What applies to this server:
What is collected. The text or file you check leaves your machine and goes to proofread.law's own server (
PROOFREAD_API, defaulthttps://proofread.law), over HTTPS, in the request that checks it.save_briefandupdate_briefsend the brief's text and title the same way;suggest_casessends the query (an article, a sentence, or the paragraph you give it) and the court, if you name one.sign_upsends the account owner's email address and the agent name you give it. Nothing else is sent: no conversation history, no other files, no telemetry.Use and storage on proofread.law. A checked input is processed in memory and discarded when the report is returned; no copy is written to disk. The exception is opt-in: a brief saved with
save_brieforupdate_briefis stored encrypted in the user's account, with its reports and earlier versions, untildelete_briefdeletes it (permanently). No other tool saves anything. The log line per request carries the kind of input, its size, the number of citations, the tier counts and the time taken, never a citation, a party name or a word of text. With an account, proofread.law keeps the email address, the plan, a monthly count of checks and a hash of each API key.Third parties. A citation string the register cannot resolve may be looked up in CourtListener's citation API (the citation string only, never prose). Deep check (
deep: true) is opt-in: the clause before each citation (up to 700 characters) and the cited opinion go to the model judge. A USsuggest_casesquery (the beta) goes to the same model judge, with the opinions of the candidate cases. Those two are the only cases in which words you send leave proofread.law. Sign-in links go through Resend; payments through Stripe, which sees the card and proofread.law does not. Requests pass through Cloudflare's edge in transit.Retention. Inputs and reports: none, except saved briefs, which are kept until the user deletes them. Account records: until the owner asks for deletion at privacy@proofread.law.
This server. It stores nothing on disk, saved briefs included (they live in the proofread.law account, not here). It keeps the last 50 reports in memory so
render_reportcan be called with a short id; they are gone when the process exits. The API key lives in your MCP client's configuration (Claude Desktop stores the extension's key as a sensitive setting); a key fromsign_upis held in memory for the session and shown to you once so you can store it. It is sent only toPROOFREAD_API, as anAuthorization: Bearerheader. If a deep-check stream stops before every citation was judged, the tool answers with an error that says how many were checked; it never presents a partial deep check as a finished one.Contact. Data Alchemy Labs, privacy@proofread.law.
The coverage caveat
Every result starts with the coverage statement, for example:
Checked against 10.1 M cases (CourtListener bulk data 2026-06-30, last refreshed 2026-09-19); federal appellate 2019 to 2023 is 10 to 15% incomplete; Westlaw/Lexis identifiers are not resolvable; statutes, regulations and secondary sources are not checked.
Read it. A citation that is not in the register is a register fact with a coverage qualifier, not proof that the case does not exist. A red row says "check this"; the tools say what was checked and what was found, never that a case is invented.
Example
Input:
Title VII forbids discrimination because of sexual orientation. Bostock v. Clayton County, 509 U.S. 644 (2020).
check_citations returns (real output, review instance, 2026-09-20):
Coverage: Checked against 10.1 M cases (CourtListener bulk data 2026-06-30, last refreshed 2026-09-19); federal appellate 2019 to 2023 is 10 to 15% incomplete; Westlaw/Lexis identifiers are not resolvable; statutes, regulations and secondary sources are not checked.
Summary: 1 citation, 1 row. Check this (red): 1. Cannot verify (orange): 0. Found (green): 0.
Flagged rows:
- CHECK THIS: 509 U.S. 644 (Bostock v. Clayton County). Register has Bostock v. Clayton County at 590 U.S. 644. Check the volume. In the register, 509 U.S. 644 is Shaw v. Reno. The case named in the document exists; this citation does not point to it. Register: https://www.courtlistener.com/opinion/4760997/bostock-v-clayton-county/
Found: 0 rows resolved to a case in the register.
Report id: r_1ff74051 (give it to render_report for a markdown report). Elapsed: 0.03 s.
Storage: Nothing you submit is stored. The document is processed in memory and discarded when this report is returned; only counts (citations, tiers, timing) are logged, never text.Development
npm install
npm run build # tsc -> dist/
npm test # vitest, mocked fetch, no network
LIVE=1 npm test -- test/live.test.ts # three live calls against proofread.law (counts against the free tier)
node scripts/smoke-stdio.mjs # spawn the stdio server, initialize, tools/list, four live tool calls
node scripts/smoke-stdio.mjs --offline # the same without networkLayout: src/client.ts is the typed HTTP client (/verify, /render, /api/coverage, /v1/resolve single and batch, /v1/briefs, /v1/suggest), src/format.ts the compact formatter for checks, src/resolve_format.ts the one for register answers, src/briefs_format.ts the one for saved briefs (/v1/briefs), src/suggest_format.ts the one for case suggestions (/v1/suggest), src/tools/<name>.ts one file per tool, src/server.ts registers them, src/cli.ts picks the transport. The remaining register routes (/v1/extract, /v1/case/{id}, /v1/coverage per reporter) slot in the same way: one method on the client, one file under src/tools/.
Publishing
See RELEASE.md: npm, the MCP Registry (server.json is in the repository), the Claude Desktop extension bundle (manifest.json, scripts/build-mcpb.sh, attached to each GitHub release) and Anthropic's connector directory.
License
MIT, Data Alchemy Labs.
Available Tools
14 toolsbilling_linkGet a checkout link for a paid planAIdempotent
Get a Stripe Checkout link that upgrades this API key's account to a paid plan (POST /agent/checkout-link). Use it when a check answers 'needs the ... plan' (402) or 'monthly allowance used' (429), or when the user asks to upgrade. Needs an API key (PROOFREAD_API_KEY, or one from sign_up). Plans: payg = pay as you go (no monthly fee; per check and per deep-checked citation), solo = monthly, firm = monthly with more keys; current prices at https://proofread.law/pricing. Give the link to the account owner to open in a browser; the plan is live within a minute of payment. No charge happens until the owner pays. An account that already has a subscription answers 409 (the owner changes plans in the billing portal at https://proofread.law/account).
Inputs: plan.
check_citationsCheck legal citations in textARead-onlyIdempotent
Check every case citation in a text against proofread.law's register of about 10 million US court opinions (CourtListener bulk data). Use it on a draft brief, memo, letter or any prose that cites cases, before the citations are relied on. Returns the coverage statement, counts per tier, one line per row that needs a human (red = check this: the register holds something concrete that disagrees, such as a different case at that citation or quoted words not in the opinion; orange = cannot verify: nothing to check against, such as a Westlaw/Lexis identifier or a volume newer than the register), the number of citations found, and a report id for render_report. Cannot: resolve Westlaw (WL) or Lexis identifiers, check statutes, regulations or secondary sources, or say whether a case is still good law. A red row means 'check this', never 'this case does not exist'; an orange row means the register has nothing to check against, which is not evidence either way. deep=true also asks, for each found citation, whether the opinion supports the sentence it is cited for (white rows). It is slower (1 to 2 s per citation), opt-in because the clause before each citation is sent to a model judge, limited to 3 per month on the free tier, and its answers are a review queue, not a verdict. Free tier: 20 checks a month. Without an API key the free tier applies per IP address; a key from sign_up (free tier) identifies the account, and a paid-plan key lifts the limits.
Inputs: text (required), deep.
check_documentCheck legal citations in a fileARead-onlyIdempotent
Check every case citation in a document on disk (PDF, DOCX, TXT or Markdown, up to 10 MB) against proofread.law's register of about 10 million US court opinions. The file is read here and uploaded to proofread.law, which extracts the text in memory, checks it and discards it. Returns the same compact result as check_citations: coverage statement, counts per tier, one line per red (check this) or orange (cannot verify) row, the number of citations found, and a report id for render_report. Scanned PDFs without a text layer, encrypted PDFs and legacy .doc files cannot be read; .docx needs a paid plan. Cannot: resolve Westlaw (WL) or Lexis identifiers, check statutes, regulations or secondary sources, or say whether a case is still good law. A red row means 'check this', never 'this case does not exist'; an orange row means the register has nothing to check against, which is not evidence either way. deep=true also asks, for each found citation, whether the opinion supports the sentence it is cited for (white rows). It is slower (1 to 2 s per citation), opt-in because the clause before each citation is sent to a model judge, limited to 3 per month on the free tier, and its answers are a review queue, not a verdict. Without an API key the free tier applies per IP address; a key from sign_up (free tier) identifies the account, and a paid-plan key lifts the limits.
Inputs: path (required), deep.
coverageWhat proofread.law checks againstARead-onlyIdempotent
The coverage statement (which opinions the register holds, its date, its known gaps, what is not checked: Westlaw/Lexis identifiers, statutes, regulations, secondary sources) and the storage notice (a checked text is not stored; only a brief the user saves with save_brief is kept, encrypted in their account). Pass jurisdiction 'ch' for the Swiss register instead (BGE/ATF/DTF, Federal Supreme Court dockets, the federal courts and the 26 cantons), which also lists the courts held and the share of the live index each covers. Call it when a user asks what the check covers, how current it is, or what happens to their text. Free, not counted as a check.
Inputs: jurisdiction.
delete_briefPermanently delete a saved briefADestructiveIdempotent
Permanently delete a brief and all its saved versions from the user's proofread.law account. The deletion is permanent: it cannot be undone and the text cannot be recovered afterwards. Only call it when the user asks to delete that brief; if the id is in any doubt, confirm it with list_briefs first. Deleting runs no check. Saving is opt-in: nothing is saved unless save_brief or update_brief is called (check_citations and check_document never save anything). A saved brief is stored encrypted in the user's proofread.law account until delete_brief removes it, and needs an API key (PROOFREAD_API_KEY, or one from sign_up).
Inputs: id (required).
get_briefGet a saved brief and its latest reportARead-onlyIdempotent
Get one brief saved in the user's proofread.law account: its saved text, the latest report (coverage statement, counts per tier, the rows that need attention, a report id for render_report) and the versions kept. Use it to pick up where the user left off, or to get the exact saved text before editing it for update_brief. include_text=false returns the report without the text (a long brief is long). Pass version to read an earlier version's text and counts instead of the latest. It only reads: this tool saves nothing and runs no check. Saving is opt-in: nothing is saved unless save_brief or update_brief is called (check_citations and check_document never save anything). A saved brief is stored encrypted in the user's proofread.law account until delete_brief removes it, and needs an API key (PROOFREAD_API_KEY, or one from sign_up).
Inputs: id (required), version, include_text.
list_briefsList the saved briefsARead-onlyIdempotent
List the briefs saved in the user's proofread.law account: id, title, when each was last saved and checked, the number of citations, the counts per tier from its latest check, and the number of versions kept. Use it to find a brief's id when the user refers to a draft by name, before get_brief, update_brief or delete_brief. It only reads: this tool saves nothing and runs no check. Saving is opt-in: nothing is saved unless save_brief or update_brief is called (check_citations and check_document never save anything). A saved brief is stored encrypted in the user's proofread.law account until delete_brief removes it, and needs an API key (PROOFREAD_API_KEY, or one from sign_up).
Inputs: no named parameters.
render_reportRender a check as a markdown reportARead-onlyIdempotent
Turn a finished check into a markdown diligence report: header, coverage and storage notices, a summary table sorted check-this, cannot-verify, support, found, and a detail block per flagged row with the register evidence. Use it when the user wants a report to keep or attach to the file, or when the compact result was cut short. Pass the report_id returned by check_citations or check_document (ids live in this server's memory until it exits), or the full report JSON from proofread.law's /verify endpoint. Free, not counted.
Inputs: report, report_id.
resolve_citationLook up one citation in the registerARead-onlyIdempotent
Look up a single case citation (for example '590 U.S. 644', or 'Bostock v. Clayton County, 590 U.S. 644 (2020)') in proofread.law's register and answer: is there a case at this citation, which one (name, court, date, parallel citations, link), and how complete the register is for that volume. This is a register lookup of the citation, not a comparison with the case name you have: if the case it returns is not the one you expected, the citation points elsewhere. Statuses: found; ambiguous (several entries, candidates listed); not in the register (a register fact with a coverage qualifier, never proof that the case does not exist); cannot verify (a Westlaw/Lexis identifier, or a volume the register cannot see yet); known citation (other opinions cite it, the opinion itself is not held); no citation recognised. Use it when one citation is in doubt; use check_citations for prose, and resolve_citations for a list. Cannot: resolve Westlaw (WL) or Lexis identifiers, check statutes, regulations or secondary sources, or say whether a case is still good law. A red row means 'check this', never 'this case does not exist'; an orange row means the register has nothing to check against, which is not evidence either way. Counts against the resolve quota (1,000 a month free), not the check quota. Without an API key the free tier applies per IP address; a key from sign_up (free tier) identifies the account, and a paid-plan key lifts the limits.
Inputs: citation (required).
resolve_citationsLook up a list of citations in the registerARead-onlyIdempotent
Look up up to 500 case citation strings in proofread.law's register in one call and get one line per citation, in input order: found (the case, court, date, link), ambiguous, not in the register (a register fact with a coverage qualifier, never proof that the case does not exist), cannot verify (Westlaw/Lexis identifier, or a volume the register cannot see yet), known citation, or no citation recognised. Use it for a table of authorities or any list of citations you already have; use check_citations for prose (it also checks names and quotations). Cannot: resolve Westlaw (WL) or Lexis identifiers, check statutes, regulations or secondary sources, or say whether a case is still good law. A red row means 'check this', never 'this case does not exist'; an orange row means the register has nothing to check against, which is not evidence either way. Each citation counts against the resolve quota (1,000 a month free), not the check quota. Without an API key the free tier applies per IP address; a key from sign_up (free tier) identifies the account, and a paid-plan key lifts the limits.
Inputs: cites (required).
save_briefSave a brief to the account and check itA
Save a brief (or any draft that cites cases) to the user's proofread.law account and check its citations in the same call. Use it when the user asks to keep a draft and come back to it, or to start the edit-and-recheck loop: save once, then after each round of edits call update_brief with the id. Returns the brief id, the coverage statement, counts per tier, one line per row that needs attention (red = check this: the register holds something concrete that disagrees; orange = cannot verify: nothing to check against, not evidence either way), and a report id for render_report. Only call it when the user wants the draft saved; to check without saving, use check_citations. Each call saves a new brief and counts as one check. Cannot: resolve Westlaw (WL) or Lexis identifiers, check statutes, regulations or secondary sources, or say whether a case is still good law. A red row means 'check this', never 'this case does not exist'; an orange row means the register has nothing to check against, which is not evidence either way. Saving is opt-in: nothing is saved unless save_brief or update_brief is called (check_citations and check_document never save anything). A saved brief is stored encrypted in the user's proofread.law account until delete_brief removes it, and needs an API key (PROOFREAD_API_KEY, or one from sign_up).
Inputs: text (required), title.
sign_upCreate a proofread.law account and API keyA
Open a proofread.law account for its owner and get an API key, in one call (POST /agent/signup). Use it when the user wants their own quota instead of the anonymous per-IP free tier, or before billing_link. The email must be the account OWNER's real inbox (placeholder domains are rejected); the owner receives one confirmation email and nothing else. The key is shown once: this server adopts it for the rest of this session, and the user should put it in PROOFREAD_API_KEY in their MCP configuration so it survives a restart. Never send the key anywhere but proofread.law. Calling again with the same address while the account is unconfirmed and unpaid rotates the key; a confirmed or paying account answers 409 (the owner manages keys at https://proofread.law/account). Free tier per account: 20 checks, 3 deep checks, 1,000 resolves a month; 5 sign-ups an hour per client.
Inputs: email (required), agent_name (required).
suggest_casesSuggest leading Swiss cases for a statute articleARead-onlyIdempotent
Swiss law only. Give a Swiss federal statute article ('Art. 41 OR', 'art. 41 CO', 'Art. 8 ZGB', 'art. 9 Cst.') or a paragraph that cites one, and get the leading Federal Supreme Court cases (BGE/ATF/DTF) the court cites with that article, as cases to read. The query needs a statute article: there is no free-text search, and a query without one answers that no article was recognised. Federal acts only; cantonal law is not covered. The answer is one ranked list, each case labelled with its field of law; the domain filter keeps the same order within one field. Each row has the citation, date, field, rank, the quoted passage (regeste or consideration), the decision's link, a link to check the citation on proofread.law, and any later change of practice (a changed precedent is listed with its flag, never left out; a row without a flag is not evidence that its practice still holds). A suggestion has not been checked against the user's sentence; to check a citation, use check_citations. The answer comes back in the query's language (German, French or Italian; lang='en' for English). During the trial phase it needs a paid-plan or trial API key; each answered query counts as one resolve.
Inputs: query (required), k, lang, domain.
update_briefSave an edited brief and re-check itA
Save an edited version of a brief saved in the user's proofread.law account and re-check it. This is the loop for fixing flagged citations: edit the text (the citation, case name or quotation a row points at, or take the citation out), call update_brief with the whole edited text, and read the changes: which flags were resolved (flagged before, not flagged now), which flags are new, how many rows are unchanged, and then every row that still needs attention. Repeat until every remaining row has been reviewed by the user. Send the full text, not a diff or an excerpt: it becomes the latest version, and the previous version is kept (get_brief lists the versions). recheck=true re-runs the check on the saved text without editing it, e.g. after the register was updated. A changed text or a recheck counts as one check and keeps a new version; a title alone renames the brief without a check and is not counted. Cannot: resolve Westlaw (WL) or Lexis identifiers, check statutes, regulations or secondary sources, or say whether a case is still good law. A red row means 'check this', never 'this case does not exist'; an orange row means the register has nothing to check against, which is not evidence either way. Saving is opt-in: nothing is saved unless save_brief or update_brief is called (check_citations and check_document never save anything). A saved brief is stored encrypted in the user's proofread.law account until delete_brief removes it, and needs an API key (PROOFREAD_API_KEY, or one from sign_up).
Inputs: id (required), text, title, recheck.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 14 tools
Tools are mostly distinct: check_citations vs check_document differ by input type (text vs file), resolve_citation vs resolve_citations differ by count, and suggest_cases offers a different domain. The cluster of citation tools could be confused, but descriptions clearly separate prose checking, list resolution, single lookup, and Swiss suggestions. Brief CRUD and account tools are unambiguous.
Most tools follow a verb_noun pattern (check_citations, resolve_citations, list_briefs, save_brief, render_report), with a few exceptions like 'coverage' (noun) and 'sign_up' (phrasal verb). The pattern is predictable and readable, with only minor deviations.
14 tools is well within the ideal range for this domain. Each tool has a clear purpose: checking, resolving, saving, managing, reporting, and account setup. No redundancy or bloat; the count matches the feature set.
The tool set covers the full lifecycle: checking citations in text and files, resolving single and batch citations, suggesting Swiss cases, managing saved briefs (CRUD), generating reports, and handling account quotas. Minor gaps exist, like no direct tool to check statutes or regulations, but those are explicitly out of scope, so the surface is appropriately complete.
Maintenance
Related MCP Connectors
Checks every citation in a US legal filing: cases, statutes, quotations and links, before you sign.
Verify legal citations, case treatment, quotes and whole briefs against 10.7M U.S. opinions
Verify legal citations and search case law and legislation from your AI assistant.
Resolve, search and verify legal citations against the official sources, with provenance.
Related MCP Servers
- FlicenseNot gradedqualityDmaintenanceEnables US case law search, citation parsing, practice management via Clio, and federal court filings through PACER.-
- AlicenseNot gradedqualityDmaintenanceValidates legal citations against the CourtListener database to detect hallucinated citations in legal documents.5MIT
- FlicenseNot gradedqualityDmaintenanceDeterministic legal citation verification for AI-generated legal briefs. Three-layer verification: CourtListener database lookup, quote-match against primary source, and LLM edge-case verification. Free tier available.-
- AlicenseAqualityCmaintenanceEnables Claude Desktop and Claude Code users to get AI-powered legal risk classifications for small business decisions, returning structured GREEN/YELLOW/RED verdicts with statute citations and compliant alternatives.1MIT