@pipeworx/mcp-emarket-storage 0.1.0

package/src/index.ts

@@ -0,0 +1,1183 @@
+interface McpToolDefinition {
+  name: string;
+  description: string;
+  /** Human-facing one-liner (fleet #1967). Optional; consumers fall back to
+   *  description. Kept in step with shared/src/types.ts — scripts/lib/
+   *  check-inlined-types.mjs reports drift at publish time. */
+  summary?: string;
+  inputSchema: {
+    type: 'object';
+    properties: Record<string, unknown>;
+    required?: string[];
+    anyOf?: Array<{ required: string[] }>;
+    oneOf?: Array<{ required: string[] }>;
+    allOf?: Array<{ required: string[] }>;
+  };
+  outputSchema?: Record<string, unknown>;
+}
+
+interface McpToolExport {
+  tools: McpToolDefinition[];
+  callTool: (name: string, args: Record<string, unknown>) => Promise<unknown>;
+  meter?: { credits: number };
+  cost?: Record<string, unknown>;
+  provider?: string;
+}
+
+/**
+ * Was this failure OUR OWN web service? — the other half of `internal-db-class.ts`.
+ *
+ * fleet #1089 pulled failures from our own Postgres out of `upstream_down` by
+ * keying on the SQLSTATE inside PostgREST's four-key error envelope. That
+ * covered the majority and structurally could not cover the rest: the rest
+ * never reach Postgres, so they carry no SQLSTATE. What was left, measured over
+ * the 24h to 2026-09-02T15:00Z (fleet #1096):
+ *
+ *     5  pipeworx-catalog  get_pack_tools     Pipeworx catalog error: 522 — error code: 522
+ *     3  fleet             fleet_list_open …  upstream_down: Fleet task queue did not respond within 25s
+ *
+ * 521/522/523/526 are Cloudflare saying its edge could not reach an ORIGIN, and
+ * in both of those rows the origin is ours — `gateway.pipeworx.io` for the
+ * catalog pack (it self-fetches when the gateway hasn't injected a manifest),
+ * our own Supabase for fleet. There is no third party anywhere in either call.
+ * Same defect as #1089: our own outage filed under `upstream_down`, the one
+ * class that means "the source is unreachable and there is nothing for us to
+ * fix", which is why the problem-tools triage skips it.
+ *
+ * WHY NOT A WORDING RULE. The obvious fix is to match `fleet db error:` and
+ * `Pipeworx catalog error:` in classifyToolError. Each is emitted from exactly
+ * one site today, so it would work today. It would also rot the first time
+ * somebody rewords a label — silently, and in the direction of hiding our own
+ * outage, which is worse than the bug being fixed. Every prose rule in
+ * error-class.ts has needed widening as packs invented new wording (#409/#450/
+ * #584); that history is most of that file's comment budget.
+ *
+ * WHAT THIS KEYS ON INSTEAD: **the host the call actually reached.** A URL's
+ * hostname is a fact about the call, not a guess about its prose. Two
+ * consequences that a pack-level flag could not give us, and the reason the
+ * flag was rejected:
+ *
+ *   - It describes the CALL, not the pack. `govcon-intel` fans out to our own
+ *     Supabase AND to genuine third parties; `court-listener` holds our cache
+ *     in Supabase and fetches courtlistener.com. An `internallyHosted: true` on
+ *     either pack would relabel a real third-party outage as ours — inventing
+ *     work, which is the same class of error in the opposite direction.
+ *   - It covers every future internal pack for free, instead of one declared
+ *     slug at a time.
+ *
+ * WHY IT SURVIVES A REWORD. The marker below is not matched as a literal by two
+ * separate files. `markInternalOrigin()` writes it and `internalHostMetricsClass()`
+ * reads it, both from the single exported `INTERNAL_ORIGIN_MARKER` constant in
+ * this module — so changing the wording changes both sides in the same edit and
+ * cannot desynchronise them. The pack's own label (`fleet db error:`,
+ * `Pipeworx catalog error:`) is not read at all: reword it freely, the class is
+ * unaffected. That is the property `stripClassPrefix` lacked when it drifted
+ * from its own classifier three times and needed a CI gate to hold them
+ * together.
+ *
+ * WHERE THE 5xx TEST LIVES. `markInternalOrigin` is called from the places that
+ * hold the real `Response` — `httpError`/`httpErrorMessage` and the timeout
+ * branch of `fetchWithTimeout` in `shared/src/http.ts` — so "is this an
+ * availability failure" is decided from the actual status code, never re-derived
+ * by scraping a number out of a sentence. A 404 from our own registry for a slug
+ * that does not exist is a caller's bad argument and is deliberately NOT marked.
+ */
+
+/**
+ * OUR OWN web service was unreachable — not an upstream, and never `upstream_down`.
+ *
+ * ONE value, not three, unlike `internal_db_*`. That split existed because a
+ * slow query, an exhausted pool and an unknown SQLSTATE have different owners
+ * and different fixes. Here there is only one story to tell — an origin we run
+ * did not answer the edge — and one owner. A bucket with no distinct owner per
+ * value is decoration; #724 is what happens when a class holds several
+ * situations, and inventing sub-values ahead of a reason to act on them
+ * differently is the same mistake with the sign flipped.
+ *
+ * METRICS ONLY, exactly like PLATFORM_KEY_ERROR_CLASS and the internal_db
+ * values. `classifyToolError` still answers `upstream_down` for the retry and
+ * hint paths, which only care whether retrying or a sibling tool might work —
+ * and it might. Nothing a caller sees or is charged changes here.
+ *
+ * READ SIDE: this value is in BROKEN_TOOL_CLASSES, FAULT_CLASSES and
+ * ALL_ERROR_CLASSES in `workers/registry-api/src/index.ts`. All three, or it
+ * lands on no dashboard — fleet #721 is the warning, where the #719 split
+ * worked on the write side and was invisible for weeks.
+ */
+const INTERNAL_SERVICE_UNREACHABLE_CLASS = 'internal_service_unreachable';
+
+/**
+ * The token that carries "this origin is ours" from the call site to the
+ * classifier.
+ *
+ * Appended to the error message rather than attached to the Error object,
+ * because the object does not survive the trip: 275 packs return `{ error:
+ * string }` instead of throwing, the gateway reads `observedError` as a string,
+ * and the fleet pack rebuilds its error from a captured status + body across a
+ * retry loop. A property on an Error would be dropped by every one of those
+ * paths and the class would work in tests and vanish in production.
+ *
+ * WORDING IS LOAD-BEARING, same rule as labelAge's note in authority.ts. This
+ * string is appended to a pack's thrown Error message (shared/src/http.ts),
+ * and a thrown Error's message is exactly what the gateway hands back to the
+ * caller as `content[0].text` when nothing rewrites it (workers/gateway/src
+ * catches the throw and sets `rawResult.message = stripClassPrefix(error)`,
+ * which does not touch this suffix) — so the original wording,
+ * " [pipeworx-hosted origin — our own service, not a third party]", was not a
+ * theoretical leak: it shipped live on pipeworx-catalog's 522s, 7 times in 6
+ * hours on 2026-09-02 (see tests/golden-internal-service.test.ts), verbatim
+ * naming Pipeworx as the host. check:hosting-claims never caught it because it
+ * did not scan shared/ at all (task #2009). Reworded to describe the
+ * OBSERVATION (the origin did not answer) without a claim about who runs it —
+ * the identical fix labelAge got: drop the possessive, keep the fact.
+ */
+const INTERNAL_ORIGIN_MARKER = ' [origin did not respond — retry before concluding the named source is down]';
+
+/**
+ * Supabase's data plane for a project is `<ref>.supabase.co`, where the ref is
+ * exactly twenty lowercase letters (ours is `pqauisounztsgdgfkhke`).
+ *
+ * Matching the shape rather than listing the ref keeps this correct when we add
+ * a project — `supabaseEnv` on a pack entry already points some packs at a
+ * second one — while still excluding `status.supabase.co`, which is Supabase's
+ * own status page and emphatically not our database. Verified 2026-09-02 by
+ * `grep -rhoE '[a-z0-9-]+\.supabase\.(co|in)' mcps shared workers scripts`: the
+ * only real project ref anywhere in the tree is ours, the rest are doc
+ * placeholders (`abc`, `xyz`, `example`) which this pattern also excludes. Same
+ * finding internal-db-class.ts relies on for the PostgREST envelope being ours
+ * by construction.
+ */
+const SUPABASE_PROJECT_HOST = /^[a-z]{20}\.supabase\.(co|in)$/;
+
+/**
+ * Is this a host WE run?
+ *
+ * Deliberately NOT including `*.workers.dev`: plenty of third-party APIs are
+ * hosted on workers.dev, so the suffix says where something runs and not who
+ * owns it. Every internal call we actually make goes to a `pipeworx.io`
+ * hostname or to our Supabase project, both of which are ownership facts.
+ *
+ * `workers/gateway/src/provenance.ts`'s `OUR_HOSTS` answers the same
+ * question and DOES include `workers.dev` — a documented divergence
+ * (task #2051), not a bug to converge. That list decides what a response may
+ * cite as a data SOURCE, where a false negative (citing our own worker as an
+ * external source) is the hosting-disclosure leak this whole file exists to
+ * prevent, so it errs broad. This one decides who gets BLAMED for a 5xx in
+ * outage metrics read by on-call, where a false positive (crediting our own
+ * infra with a third party's outage) hides the real failure, so it errs
+ * narrow. Same suffix, opposite direction, because they are never called for
+ * the same reason.
+ *
+ * Returns false on anything unparseable rather than throwing — this runs inside
+ * an error path, and an error path that can itself throw turns a diagnosable
+ * failure into a mystery.
+ */
+function isPipeworxOrigin(url: string | URL | undefined | null): boolean {
+  if (!url) return false;
+  let host: string;
+  try {
+    host = new URL(url instanceof URL ? url.href : url).hostname.toLowerCase();
+  } catch {
+    return false;
+  }
+  if (host === 'pipeworx.io' || host.endsWith('.pipeworx.io')) return true;
+  return SUPABASE_PROJECT_HOST.test(host);
+}
+
+/**
+ * Append the marker when this failure was OUR origin failing to answer.
+ *
+ * `status` is the HTTP status when there is one, and omitted for a timeout —
+ * where there is no response at all, and "the origin did not answer" is the
+ * whole observation. Statuses below 500 are left alone: a 404 from our own
+ * registry for a slug that does not exist is the caller's argument, not our
+ * outage, and marking it would put ordinary 404s on the incident dashboard.
+ *
+ * Idempotent, so a message that is wrapped and re-marked on the way up (the
+ * fleet pack's retry loop re-throws through two layers) carries the marker once.
+ */
+function markInternalOrigin(
+  message: string,
+  url: string | URL | undefined | null,
+  status?: number,
+): string {
+  if (status !== undefined && status < 500) return message;
+  if (!isPipeworxOrigin(url)) return message;
+  if (message.includes(INTERNAL_ORIGIN_MARKER)) return message;
+  return message + INTERNAL_ORIGIN_MARKER;
+}
+
+/**
+ * Which blob4 value a failure from our own web services books as, or undefined
+ * if this is not one.
+ *
+ * Ordered AFTER `internalDbMetricsClass` at the call site: a PostgREST envelope
+ * from our own Supabase is a strictly more specific statement about the same
+ * row (which of our services, and why), and the two cannot disagree about
+ * whether the failure is ours.
+ */
+function internalHostMetricsClass(error: string): string | undefined {
+  return error.includes(INTERNAL_ORIGIN_MARKER) ? INTERNAL_SERVICE_UNREACHABLE_CLASS : undefined;
+}
+
+
+/**
+ * One place to turn a failed `fetch` into an error a caller can act on.
+ *
+ * Nearly every pack was written the same way:
+ *
+ *     if (!res.ok) throw new Error(`Unsplash: ${res.status}`);
+ *
+ * which discards the response body — and the body is usually where the upstream
+ * says what was actually wrong ("**symbol** not found: GBP", "parameter `year`
+ * out of range", "unknown taxonomy id"). The caller gets a number, cannot
+ * self-correct, and retries the same broken call. A 2026-07-31 sweep found this
+ * shape in 481 of 1,400 packs, 47 of them PLATFORM-keyed.
+ *
+ * It also hides bugs one level down. Two of the first three packs audited had a
+ * second defect that only existed because of this line: unsplash's rate-limit
+ * branch sat BELOW a catch-all and was unreachable, and bea-gov parsed
+ * `BEAAPI.Error.APIErrorDescription` below a `!res.ok` throw that made the
+ * parsing dead code for every non-200.
+ *
+ * DELIBERATELY NOT A CLASSIFIER. It does not add `user_error:` /
+ * `upstream_down:` prefixes. Those decide which tier a failure lands in, and the
+ * `error` tier is what the daily problem-tools list is built from — it means
+ * "Pipeworx has a defect". A 400 is genuinely ambiguous: often a caller's bad
+ * argument, but sometimes a query WE built wrong (ted-eu comma-joined its CPV
+ * values into something TED rejected, and that bug was found only because it sat
+ * in `error`). Blanket-classifying 400s as caller mistakes would have hidden it.
+ * A pack that KNOWS which it is should keep saying so explicitly; this helper is
+ * for the 481 that say nothing at all.
+ */
+
+/** Longest upstream explanation we'll pass through. Enough for a real message,
+ *  short enough that an HTML page or a stack trace can't swamp the error. */
+
+const MAX_DETAIL = 300;
+
+/**
+ * Default bound for `fetchWithTimeout` when a pack doesn't state its own.
+ *
+ * 25s mirrors the number `epo-ops` landed on after measuring the real failure:
+ * a degraded upstream that doesn't error, it just never answers, and a Worker
+ * sits in `await fetch()` until ITS OWN execution budget kills the request —
+ * which can take minutes, not seconds (epo_ops_search_patents measured 4-8
+ * MINUTE hangs before this existed). 25s is short enough that a caller gets a
+ * fast, actionable error instead of holding the connection, and long enough
+ * that it doesn't false-trip on a merely-slow-but-alive upstream.
+ */
+const DEFAULT_FETCH_TIMEOUT_MS = 25_000;
+
+/**
+ * Read the body of a failed response and fold it into a throwable Error.
+ *
+ * Usage — note the `await`, which is the one thing that makes this a mechanical
+ * change rather than a drop-in:
+ *
+ *     if (!res.ok) throw await httpError(res, 'Unsplash');
+ *
+ * Safe to call on any non-ok response: a body that is missing, empty, unreadable
+ * or HTML degrades to exactly the old `Name: 404` string rather than throwing
+ * something new from inside the error path.
+ */
+async function httpError(res: Response, name: string): Promise<Error> {
+  return new Error(await httpErrorMessage(res, name));
+}
+
+/** The message text without constructing an Error — for packs that need to wrap
+ *  it in their own envelope or add an explicit classification prefix. */
+async function httpErrorMessage(res: Response, name: string): Promise<string> {
+  // The one place a 5xx from a host WE run gets stamped as ours. `res.url` is
+  // the URL the fetch actually resolved to (after redirects), so this is a fact
+  // about the call rather than a guess from the `name` the pack passed in —
+  // reword that label freely, the class does not move. See
+  // internal-host-class.ts; no-op for every third-party upstream, which is why
+  // this touches 481 packs' error text and changes none of it.
+  return markInternalOrigin(
+    `${name}: ${res.status}${detailSuffix(await readDetail(res))}`,
+    res.url,
+    res.status,
+  );
+}
+
+/**
+ * Just the upstream's own explanation — no name, no status.
+ *
+ * For a pack that has already said both in its own sentence. epo-ops reads
+ * `EPO rejected this search as too large (HTTP 413) — ${httpErrorMessage(…)}`,
+ * which rendered as `… (HTTP 413) — EPO: 413.` once the XML detail was being
+ * dropped: the upstream named twice, the status twice, and the one thing EPO
+ * actually said ("Not enough characters before truncation character") nowhere
+ * (fleet #712). Returns '' when the body carries nothing readable, so a caller
+ * can fall back to its own wording.
+ */
+async function upstreamDetail(res: Response): Promise<string> {
+  return readDetail(res);
+}
+
+/**
+ * Read a SUCCESSFUL response as JSON, failing loudly when it isn't JSON.
+ *
+ * `httpError` above only ever runs on `!res.ok`, which leaves the nastier half
+ * of the problem unhandled: an upstream that answers **HTTP 200 with an HTML
+ * page**. A bot wall, a login redirect, a maintenance interstitial and a CDN
+ * error page are all 200s, so `res.ok` is true, and `res.json()` then throws
+ * `Unexpected token '<', "<!DOCTYPE "... is not valid JSON`.
+ *
+ * That string is the problem. It names no upstream, carries no status, and
+ * reads like a parser bug in Pipeworx — so it lands in the `error` tier, which
+ * means "we have a defect", and the caller is told nothing they can act on.
+ * data.govt.nz sat dead behind an Imperva challenge this way and every
+ * status-code health check we own reported it green (7889a845). A zero-length
+ * body has the same shape: `Unexpected end of JSON input`, seen this week on
+ * uk-gazette (83% of external calls) and census.
+ *
+ * UNLIKE `httpError`, this one DOES classify, and the asymmetry is deliberate.
+ * A 400 is genuinely ambiguous — often the caller's bad argument, sometimes a
+ * query we built wrong — so blanket-classifying it would hide our own bugs.
+ * There is no such ambiguity here: **no argument a caller can pass makes a JSON
+ * API return an HTML page.** It is always the upstream, so `upstream_down:` is
+ * a statement of fact rather than a guess, and it keeps these out of the
+ * problem-tools list where they crowd out real defects.
+ *
+ *     const data = await parseJson<Feed>(res, 'UK Gazette');
+ *
+ * Call it only after the `!res.ok` check — on a failed response you want
+ * `httpError`, which mines the body for the upstream's own explanation.
+ */
+async function parseJson<T>(res: Response, name: string): Promise<T> {
+  let raw: string;
+  try {
+    raw = await res.text();
+  } catch {
+    throw new Error(
+      `upstream_down: ${name} returned a body that could not be read (HTTP ${res.status}). ` +
+        'The connection most likely dropped mid-response; retrying is reasonable.',
+    );
+  }
+
+  const type = res.headers.get('content-type') ?? 'no content-type';
+
+  if (!raw.trim()) {
+    throw new Error(
+      `upstream_down: ${name} answered HTTP ${res.status} with an EMPTY body where JSON was expected (${type}). ` +
+        'Nothing about the request can cause this — it is an upstream fault, and the same call may well work on retry.',
+    );
+  }
+
+  // Checked before parsing rather than in the catch, because knowing it is
+  // markup is what turns "we failed to parse something" into "they served a
+  // web page" — the second is diagnosable, the first is not.
+  const head = raw.slice(0, 200).trimStart().toLowerCase();
+  if (head.startsWith('<!doctype') || head.startsWith('<html') || head.startsWith('<?xml')) {
+    const kind = head.startsWith('<?xml') ? 'an XML document' : 'an HTML page';
+    // The summary, not the source. Pasting the first 120 characters of a web
+    // page handed the agent `<!DOCTYPE html><html lang="en"…` — the same leak
+    // this branch exists to describe (fleet #712).
+    throw new Error(
+      `upstream_down: ${name} answered HTTP ${res.status} with ${kind} instead of JSON (${type}). ` +
+        'That is typically a bot wall, a login redirect or a maintenance page — it is returned as a SUCCESS, ' +
+        `so status-code health checks read it as fine. No argument change will get past it. ` +
+        `The page says: ${summarizeErrorBody(raw) || 'nothing readable'}`,
+    );
+  }
+
+  try {
+    return JSON.parse(raw) as T;
+  } catch {
+    throw new Error(
+      `upstream_down: ${name} answered HTTP ${res.status} with a body that is not valid JSON (${type}). ` +
+        `It begins: ${stripMarkup(raw).slice(0, 120) || '(unreadable)'}`,
+    );
+  }
+}
+
+/**
+ * `fetch`, but bounded — the fix for a systemic gap found 2026-08-30: a grep
+ * audit of every pack's `mcps/*\/src/index.ts` found 1,339 of ~1,500 call
+ * `fetch()` with NO timeout guard anywhere in the file. Two of those
+ * (epo-ops, statcan) were confirmed live-hanging for 4-8 minutes before this
+ * existed — every unguarded call carries the same risk, just unconfirmed.
+ *
+ * Mirrors the `epoFetch` wrapper `mcps/epo-ops/src/index.ts` shipped first:
+ * bound the request with `AbortSignal.timeout`, and on a timeout/abort throw
+ * an `upstream_down:` error that names the upstream and the bound rather than
+ * letting the raw `TimeoutError`/`AbortError` (which names neither) propagate.
+ * `upstream_down:` is deliberate, same reasoning as `parseJson` above — no
+ * argument a caller passes can make an upstream hang, so it is always the
+ * upstream's fault, and marking it that way keeps a slow API off the
+ * problem-tools list where it would crowd out our own defects.
+ *
+ * Usage — a mechanical swap for a bare `fetch(url, init)`:
+ *
+ *     const res = await fetchWithTimeout(url, init, 'Some API');
+ *
+ * Pass `timeoutMs` as a fourth argument to override the default for a pack
+ * with a known-slower upstream; the label should be the same short name you'd
+ * pass to `httpError`/`httpErrorMessage` for that call.
+ */
+async function fetchWithTimeout(
+  url: string | URL,
+  init: RequestInit = {},
+  name: string,
+  timeoutMs: number = DEFAULT_FETCH_TIMEOUT_MS,
+): Promise<Response> {
+  try {
+    return await fetch(url, { ...init, signal: AbortSignal.timeout(timeoutMs) });
+  } catch (err) {
+    if (err instanceof Error && (err.name === 'TimeoutError' || err.name === 'AbortError')) {
+      // States the OBSERVATION (no response in N seconds), not a diagnosis.
+      // "appears to be degraded" is an inference about the vendor that we have
+      // not checked, and it is wrong in a way that misdirects whoever reads it:
+      // a timeout from a Worker can equally mean OUR egress is blocked.
+      //
+      // Measured today (2026-09-01, fleet #1047): every call to
+      // mainnet.base.org failed from the x402 facilitator while the identical
+      // request from a laptop returned 200. Base was entirely healthy; the
+      // public RPC refuses Cloudflare Worker egress. Had this message fired
+      // there it would have blamed Base by name, and the next person would have
+      // waited for a vendor outage to clear that did not exist.
+      // A timeout has no status to test — there is no response at all — so
+      // `markInternalOrigin` is called without one: an origin we run that never
+      // answered is an availability failure by definition. This is the half of
+      // fleet #1096 with neither a SQLSTATE nor a status code to key on.
+      throw new Error(
+        markInternalOrigin(
+          `upstream_down: ${name} did not respond within ${timeoutMs / 1000}s. ` +
+            `That can be ${name} being slow or down, or this environment being unable to reach it ` +
+            `(some hosts refuse datacenter/Worker egress) — retry shortly, and check reachability ` +
+            `from elsewhere before concluding ${name} is down.`,
+          url,
+        ),
+      );
+    }
+    throw err;
+  }
+}
+
+function detailSuffix(detail: string): string {
+  return detail ? ` — ${detail}` : '';
+}
+
+async function readDetail(res: Response): Promise<string> {
+  let raw: string;
+  try {
+    raw = await res.text();
+  } catch {
+    // Body already consumed, or the connection died mid-read. The status alone
+    // is still worth throwing — never let the error path throw its own error.
+    return '';
+  }
+  return summarizeErrorBody(raw);
+}
+
+/**
+ * Turn ANY error body — JSON, HTML, XML or plain text — into one short phrase
+ * that never contains markup.
+ *
+ * This used to just drop an HTML or XML body on the floor, on the reasoning
+ * that markup crowds out the status. That was half right. Dropping it loses the
+ * one sentence a caller could have acted on: an `Access Denied` title, an SDMX
+ * `<message:Error>` text, an OPS fault string. A 2026-08-30 support sweep
+ * measured 13 of 291 caller-facing error rows carrying a raw page or document
+ * verbatim, across 11 packs, and in every one of them the useful content —
+ * "Access Denied", "Invalid country code", "SCRAPE_TIMEOUT" — was in there,
+ * buried in markup the agent had to parse out of a string (fleet #712).
+ *
+ * So: extract the meaning, discard the markup. The output is passed through
+ * `stripMarkup` unconditionally, which is what lets `check:error-body-leak`
+ * assert mechanically that no caller-facing message can contain `<?xml`,
+ * `<!DOCTYPE` or `<html`.
+ */
+function summarizeErrorBody(raw: string): string {
+  if (!raw || !raw.trim()) return '';
+
+  const head = raw.slice(0, 400).trimStart().toLowerCase();
+
+  // An HTML error page (Cloudflare interstitial, nginx default, a login
+  // redirect) says what it is in its <title>, and almost nowhere else.
+  if (head.startsWith('<!doctype') || head.startsWith('<html')) {
+    const title = htmlTitle(raw);
+    return title
+      ? `${title} (upstream returned an HTML error page, not an API response)`
+      : 'upstream returned an HTML error page, not an API response';
+  }
+
+  // XML fault documents — EPO OPS, SDMX (`<message:Error>`), SOAP faults. The
+  // human sentence sits in a child element whose tag name says what it is.
+  if (head.startsWith('<?xml') || head.startsWith('<')) {
+    const fault = xmlFaultText(raw);
+    return fault
+      ? `${stripMarkup(fault).slice(0, MAX_DETAIL)} (from the upstream's XML error document)`
+      : 'upstream returned an XML error document with no readable message';
+  }
+
+  // Most JSON error bodies bury one human sentence among ids and echoed request
+  // params. Prefer that sentence; fall back to the whole body when the shape is
+  // unfamiliar, since an unfamiliar shape is exactly when we can least afford to
+  // guess wrong and show nothing.
+  const fromJson = messageFromJson(raw);
+  return stripMarkup(fromJson ?? raw).slice(0, MAX_DETAIL);
+}
+
+/** The `<title>` of an HTML error page, or its first `<h1>` — the two places a
+ *  bot wall, a 502 and an "Access Denied" all state what happened. */
+function htmlTitle(raw: string): string | null {
+  const head = raw.slice(0, 4000);
+  for (const re of [/<title[^>]*>([\s\S]*?)<\/title>/i, /<h1[^>]*>([\s\S]*?)<\/h1>/i]) {
+    const m = re.exec(head);
+    const text = m ? stripMarkup(m[1]) : '';
+    if (text) return text.slice(0, 160);
+  }
+  return null;
+}
+
+/** Tag names that carry the explanation in an XML fault document, namespace
+ *  prefix optional (`<message:Error>`, `<com:Text>`, `<faultstring>`). */
+const XML_FAULT_TAG_RE =
+  /<(?:[A-Za-z0-9_.-]+:)?(?:text|message|description|faultstring|reason|detail|title|errormessage|error)\b[^>]*>([^<]{2,400})</i;
+
+function xmlFaultText(raw: string): string | null {
+  const head = raw.slice(0, 8000);
+  const tagged = XML_FAULT_TAG_RE.exec(head);
+  if (tagged && tagged[1].trim()) return tagged[1];
+
+  // Nothing conventionally named — take the longest text node instead. A fault
+  // document with one sentence in an oddly named element is still readable;
+  // returning nothing at all is not.
+  let best = '';
+  for (const m of head.matchAll(/>([^<>]{8,400})</g)) {
+    const text = m[1].trim();
+    if (text.length > best.length) best = text;
+  }
+  return best || null;
+}
+
+/**
+ * Remove every tag and stray angle bracket, then collapse whitespace.
+ *
+ * Applied to everything on the way out, including the JSON and plain-text
+ * paths, because an upstream is free to embed markup in a JSON string field —
+ * and a leak is a leak regardless of which branch produced it.
+ */
+function stripMarkup(s: string): string {
+  return collapse(decodeEntities(s.replace(/<[^>]*>/g, ' ')).replace(/[<>]/g, ' '));
+}
+
+/** The handful of entities that show up in error-page titles. Decoded AFTER
+ *  tags are stripped and BEFORE the angle-bracket sweep, so `&lt;script&gt;`
+ *  in a title cannot decode into markup that survives — EMBL-EBI's ChEMBL 500
+ *  page renders as `500 Internal Server Error &lt; EMBL-EBI` otherwise. */
+function decodeEntities(s: string): string {
+  return s
+    .replace(/&(?:amp|#0*38);/gi, '&')
+    .replace(/&(?:lt|#0*60);/gi, '<')
+    .replace(/&(?:gt|#0*62);/gi, '>')
+    .replace(/&(?:quot|#0*34);/gi, '"')
+    .replace(/&(?:#0*39|apos|#x0*27);/gi, "'")
+    .replace(/&nbsp;/gi, ' ');
+}
+
+/** The conventional "what went wrong" field, under any of the names upstreams
+ *  actually use. Checked in order; first non-empty string wins. */
+const MESSAGE_KEYS = [
+  'message', 'error_message', 'errorMessage', 'detail', 'details',
+  'description', 'error_description', 'reason', 'title', 'fault',
+];
+
+function messageFromJson(raw: string): string | null {
+  let parsed: unknown;
+  try {
+    parsed = JSON.parse(raw);
+  } catch {
+    return null;
+  }
+  return pickMessage(parsed, 0);
+}
+
+function pickMessage(node: unknown, depth: number): string | null {
+  // Two levels covers `{error: {message}}` and `{errors: [{detail}]}`, the two
+  // shapes that account for nearly all of them, without walking a large payload.
+  if (depth > 2 || node == null) return null;
+
+  if (typeof node === 'string') return node.trim() || null;
+
+  if (Array.isArray(node)) {
+    for (const item of node) {
+      const found = pickMessage(item, depth + 1);
+      if (found) return found;
+    }
+    return null;
+  }
+
+  if (typeof node !== 'object') return null;
+  const obj = node as Record<string, unknown>;
+
+  for (const key of MESSAGE_KEYS) {
+    const v = obj[key];
+    if (typeof v === 'string' && v.trim()) return v.trim();
+  }
+  // `{error: …}` where error is itself an object or a string — the single most
+  // common wrapper, so it is worth descending into by name rather than scanning
+  // every key and risking picking up an echoed request parameter.
+  for (const key of ['error', 'errors', 'fault', 'Error', 'data']) {
+    if (key in obj) {
+      const found = pickMessage(obj[key], depth + 1);
+      if (found) return found;
+    }
+  }
+  return null;
+}
+
+/** Errors are read in a single line of log output; newlines and runs of
+ *  whitespace make a multi-line body unreadable there. */
+function collapse(s: string): string {
+  return s.replace(/\s+/g, ' ').trim();
+}
+
+
+// Reusable entity-resolution helpers for MCP packs. SELF-CONTAINED — no internal
+// imports — so publish-pack.sh can inline it into standalone pack builds the same
+// way it inlines the McpToolExport type.
+//
+// Recurring failure mode across financial packs: callers pass a company NAME
+// ("Apple", "apple inc") where a ticker / CIK / provider symbol is expected, and
+// the pack 404s or throws "not found". `rankMatches` is a generic name-ranker any
+// pack can run over its OWN list (US tickers, B3 tickers, drug names, airports…);
+// `resolveSecEntity` wraps it around the SEC company_tickers.json universe, shared
+// by the packs that key on CIK (edgar, sec).
+
+type MatchKind = 'exact' | 'prefix' | 'word' | 'substring';
+
+interface RankedMatch<T> {
+  item: T;
+  kind: MatchKind;
+  score: number;
+}
+
+const normalize = (s: string): string =>
+  s.toUpperCase().replace(/[.,]/g, '').replace(/\s+/g, ' ').trim();
+
+/**
+ * Rank `items` by how well their name matches `query`:
+ * exact (4) > prefix (3) > whole-word (2) > substring (1). Ties break by shortest
+ * name — the primary entity (e.g. "Apple Inc." over "Apple Hospitality REIT").
+ * Returns only items that match at all, best first. Pure (no I/O).
+ */
+function rankMatches<T>(
+  query: string,
+  items: T[],
+  getName: (item: T) => string,
+): RankedMatch<T>[] {
+  const q = normalize(query);
+  if (!q) return [];
+  const scored: { item: T; kind: MatchKind; score: number; len: number }[] = [];
+  for (const item of items) {
+    const name = getName(item);
+    const n = normalize(name);
+    let kind: MatchKind | null = null;
+    let score = 0;
+    if (n === q) { kind = 'exact'; score = 4; }
+    else if (n.startsWith(q)) { kind = 'prefix'; score = 3; }
+    else if (n.includes(` ${q} `) || n.endsWith(` ${q}`)) { kind = 'word'; score = 2; }
+    else if (n.includes(q)) { kind = 'substring'; score = 1; }
+    if (kind) scored.push({ item, kind, score, len: name.length });
+  }
+  scored.sort((a, b) => b.score - a.score || a.len - b.len);
+  return scored.map(({ item, kind, score }) => ({ item, kind, score }));
+}
+
+interface SecTickerRow { cik_str: number; ticker: string; title: string }
+
+interface SecEntity {
+  ticker: string;
+  cik: string;
+  cik_padded: string;
+  company_name: string;
+  matched_by: 'ticker' | 'company_name';
+  alternatives?: { ticker: string; company_name: string; cik: string }[];
+}
+
+const SEC_TICKERS_URL = 'https://www.sec.gov/files/company_tickers.json';
+
+/**
+ * Resolve a ticker OR company name to its SEC identity (CIK + canonical name).
+ * Exact ticker first (the common, unambiguous case), then fuzzy company-name
+ * fallback so "Apple" / "APPLE" → AAPL's CIK. Throws if nothing matches.
+ *
+ * `headers` lets callers pass their pack's SEC User-Agent — www.sec.gov requires
+ * a UA. `fetchImpl` defaults to global fetch (override in tests).
+ */
+async function resolveSecEntity(
+  query: string,
+  opts: { fetchImpl?: typeof fetch; headers?: Record<string, string> } = {},
+): Promise<SecEntity> {
+  if (typeof query !== 'string' || !query.trim()) {
+    throw new Error('Required argument is missing or empty. Pass a ticker like "AAPL" or a company name like "Apple".');
+  }
+  const doFetch = opts.fetchImpl ?? fetch;
+  const res = await doFetch(SEC_TICKERS_URL, { headers: opts.headers });
+  if (!res.ok) throw new Error(`SEC ticker lookup error: ${res.status}`);
+  const data = (await res.json()) as Record<string, SecTickerRow>;
+  const rows = Object.values(data);
+
+  // 1) Exact ticker match — the common, unambiguous case.
+  const q = query.toUpperCase().trim();
+  for (const r of rows) {
+    if (r.ticker === q) return toEntity(r, 'ticker');
+  }
+
+  // 2) Company-name fallback.
+  const ranked = rankMatches(query, rows, (r) => r.title);
+  if (ranked.length) {
+    const best = toEntity(ranked[0].item, 'company_name');
+    const alts = ranked.slice(1, 4).map((m) => ({
+      ticker: m.item.ticker,
+      company_name: m.item.title,
+      cik: String(m.item.cik_str),
+    }));
+    if (alts.length) best.alternatives = alts;
+    return best;
+  }
+
+  throw new Error(`No SEC company matches "${query}". Pass a US-listed ticker ("AAPL") or the exact listed-company name ("Apple Inc."). If this is a clinical-trial sponsor, an operating subsidiary (e.g. "Merck Sharp & Dohme" → Merck & Co), or a foreign/private entity, call sponsor_to_filer({sponsor}) instead — it resolves subsidiaries to the listed parent and honestly reports when no US-listed filer exists.`);
+}
+
+function toEntity(r: SecTickerRow, matched_by: 'ticker' | 'company_name'): SecEntity {
+  return {
+    ticker: r.ticker,
+    cik: String(r.cik_str),
+    cik_padded: String(r.cik_str).padStart(10, '0'),
+    company_name: r.title,
+    matched_by,
+  };
+}
+
+const GENERIC_CORP_WORDS = new Set([
+  'THE', 'A', 'INC', 'CORP', 'CORPORATION', 'CO', 'COMPANY', 'LTD', 'LIMITED',
+  'LLC', 'LP', 'PLC', 'SA', 'AG', 'NV', 'GMBH', 'AB', 'AS', 'OY', 'SPA',
+  'GROUP', 'HOLDINGS', 'HOLDING', 'AND', 'OF', 'US', 'USA', 'INTERNATIONAL',
+  'GLOBAL',
+]);
+
+/**
+ * Split a corporate/organization name into its SIGNIFICANT tokens — words
+ * that aren't generic corporate boilerplate (Inc, Co, Ltd, Group, ...) or
+ * punctuation — sorted LONGEST FIRST. Built for cross-registry name joins
+ * where the two registries anchor on different words of the same name: SEC
+ * lists Eli Lilly as "ELI LILLY & Co", but Drugs@FDA's sponsor_name field
+ * uses "LILLY" — the longer, more distinctive token, not the first one
+ * ("ELI" alone is short and matches too loosely). Pure (no I/O); callers
+ * typically try tokens in order until one call to their OWN registry
+ * returns a result.
+ */
+function significantNameTokens(name: string): string[] {
+  const tokens = name
+    .toUpperCase()
+    .replace(/[.,&/()-]/g, ' ')
+    .split(/\s+/)
+    .filter((t) => t.length > 1 && !GENERIC_CORP_WORDS.has(t));
+  return [...new Set(tokens)].sort((a, b) => b.length - a.length);
+}
+/**
+ * Italian regulated disclosures direct from eMarket Storage (Teleborsa) — keyless, same-day-fresh.
+ *
+ * The Italian sibling of amf-filings (France).
+ *
+ * eMarket Storage is one of Italy's two Officially Appointed Mechanisms
+ * (OAMs) for EU Transparency Directive regulated-information dissemination —
+ * a mandated central storage point issuers use to publish disclosures
+ * (informazioni privilegiate, notifiche di partecipazioni rilevanti,
+ * relazioni finanziarie, internal dealing, etc.).
+ *
+ * Built for the same reason as amf-filings: filings.xbrl.org (the community
+ * ESEF index we otherwise proxy) runs months behind for current-year Italian
+ * filings. This pack goes to the primary disseminator instead. Verified
+ * 2026-09-23: a category="104" (Informazioni privilegiate) query for
+ * 2026-09-22..2026-09-24 returned a WEBUILD disclosure published
+ * 2026-09-23 17:13 local time, with its PDF confirmed live (HTTP 200).
+ *
+ * There is no formal API — `/jsonapi` on this Drupal site 404s. The listing
+ * at emarketstorage.it/it/comunicati-finanziari is a plain GET-filterable
+ * Drupal View; this pack parses its HTML with targeted regex (Cloudflare
+ * Workers have no DOMParser — same approach as spain-tenders/gdacs).
+ *
+ * ⚠️ DOCUMENT INDEX, NOT XBRL FACTS. Every row is a disclosure EVENT — a
+ * title, an issuer, a timestamp, and a link to the underlying PDF. There is
+ * no per-fact structured-data extraction (no "net income = X" lookup).
+ *
+ * ⚠️ eMARKET-COVERED ISSUERS ONLY — NOT EVERY ITALIAN LISTED COMPANY. Italy
+ * has a SECOND OAM, 1INFO (a separate, keyless JSON API covering ~314
+ * issuers), and an issuer discloses through exactly one of the two. Some
+ * well-known names (e.g. IREN, SOMEC) are 1INFO issuers and will NOT be
+ * found here — confirmed 2026-09-23 by their absence from eMarket Storage's
+ * own ~480-company issuer list. This pack does not merge 1INFO; a company
+ * not found is reported as such, with that explanation, not as an error.
+ *
+ * ⚠️ NO ISIN/LEI. Unlike amf-filings, eMarket Storage's listing and its
+ * issuer picklist expose only a company NAME and an internal Teleborsa
+ * "azienda" numeric id — no ISIN or LEI field was found anywhere on the
+ * listing or a per-company page (checked 2026-09-23). Identify issuers by
+ * name (resolved against the site's own issuer list) or by azienda_id
+ * directly if you already have it from a prior search.
+ *
+ * ⚠️ date_to IS INCLUSIVE, but the upstream `data_to` param is EXCLUSIVE of
+ * the day itself — this is a silent-LOSS trap, not just a single-day one.
+ * Verified live 2026-09-24: date_from=2026-09-21&date_to=2026-09-22 against
+ * the raw source returned rows from the 21st ONLY, silently dropping the
+ * entire 22nd — a caller reading "21st through 22nd" got back a wrong
+ * answer with no error. This tool always sends upstream a bound one
+ * calendar day past the caller's inclusive date_to and echoes both values
+ * (`date_to` inclusive, `upstream_date_to_exclusive` the raw bound sent).
+ *
+ * All tools return shaped, LLM-friendly objects (not raw HTML) and never
+ * throw — fetch/parse failures resolve to { error }.
+ */
+
+
+const UA = 'Pipeworx/1.0 (+https://pipeworx.io; support@pipeworx.io)';
+const ORIGIN = 'https://www.emarketstorage.it';
+const LISTING_PATH = '/it/comunicati-finanziari';
+const PAGE_SIZE = 24; // Drupal view's fixed page size — verified live, every page but the last was exactly 24 rows.
+
+async function pwFetch(url: string | URL, init?: RequestInit): Promise<Response> {
+  return fetchWithTimeout(
+    url,
+    { ...init, headers: { Accept: 'text/html', 'User-Agent': UA, ...(init?.headers ?? {}) } },
+    'eMarket Storage (Teleborsa OAM)',
+  );
+}
+
+// EU Transparency Directive category codes, as eMarket Storage's own
+// `categoria` filter enumerates them (extracted from the site's exposed
+// filter select, 2026-09-23). Shortened to the headline + plain description;
+// the omitted middle segment is boilerplate ("tutte le informazioni
+// comunicate ai sensi dell'articolo N della direttiva 2004/109/CE").
+const CATEGORY_LABELS: Record<string, string> = {
+  '100': '1.1 Relazioni finanziarie annuali e relazioni di revisione annuali',
+  '101': '1.2 Relazioni finanziarie semestrali e revisioni limitate',
+  '102': '1.3 Pagamento ai governi',
+  '103': "2.1 Stato membro d'origine",
+  '104': '2.2 Informazioni privilegiate',
+  '105': '2.3 Notifiche di partecipazioni rilevanti',
+  '106': "2.4 Acquisizione o cessione di azioni proprie dell'emittente",
+  '107': '2.5 Totale dei diritti di voto e del capitale',
+  '108': '2.6 Modifiche dei diritti inerenti alle categorie di azioni o valori mobiliari',
+  '109': '3.1 Ulteriori informazioni previste dalla regolamentazione nazionale',
+  '110': 'MANTRA — Internal Dealing (operazioni di persone rilevanti e strettamente associate)',
+  '111': 'MANRSS — Relazione sulla gestione inclusa la dichiarazione di sostenibilità',
+  '150': "REGEM — Altre informazioni regolamentate (art. 65-ter Regolamento Consob 11971/1999)",
+};
+const CATEGORY_HINT = Object.entries(CATEGORY_LABELS).map(([code, label]) => `${code}=${label}`).join('; ');
+
+const tools: McpToolExport['tools'] = [
+  {
+    name: 'emarket_search_disclosures',
+    description:
+      `Search Italian regulated company disclosures from eMarket Storage (Teleborsa), one of Italy's two EU Transparency Directive Officially Appointed Mechanisms. PREFER OVER esef_filing_search FOR ITALIAN ISSUERS COVERED HERE — this source is updated same-day, unlike the community ESEF index which runs months behind. Identify the issuer with company (free-text name, resolved against eMarket's own ~480-issuer list) or azienda_id (Teleborsa's internal numeric id, from a prior result) — there is no ISIN/LEI field on this source. date_to IS INCLUSIVE (the last day you want included) — this tool handles the underlying source's own date_to being EXCLUSIVE by always querying one calendar day past it, so date_from="2026-09-21", date_to="2026-09-22" correctly returns rows from BOTH the 21st and the 22nd; a single date_from with no date_to searches just that one day. COVERAGE: only issuers that disclose through eMarket Storage — a second Italian mechanism, 1INFO, covers a separate set of issuers (e.g. IREN, SOMEC) and is NOT included; an unmatched company name is reported as "not an eMarket Storage issuer", not as an error. Category codes: ${CATEGORY_HINT}. Returns disclosure EVENTS (title, company, timestamp, PDF link) — NOT extracted XBRL facts.`,
+    inputSchema: {
+      type: 'object',
+      properties: {
+        company: { type: 'string', description: 'Free-text issuer name, e.g. "Saipem", "amplifon". Resolved against eMarket Storage\'s own issuer picklist (name match, not a fulltext search). If it matches more than one issuer equally well, the response lists the candidates instead of guessing — retry with azienda_id.' },
+        azienda_id: { type: ['string', 'number'], description: 'Teleborsa\'s internal issuer id, e.g. "1247" (Saipem). Use the value returned by a previous company search to skip name resolution.' },
+        category: { type: 'string', description: 'EU Transparency Directive category code to filter by, e.g. "104" for Informazioni privilegiate. See tool description for the full code list. When set, every returned row is tagged with this category (the source does not expose a category on the row itself, only as a search filter — so category is null on unfiltered results).' },
+        market: { type: 'string', description: 'Exact market/segment name to filter by, e.g. "Euronext Milan", "Euronext STAR Milan", "MTA". Must match eMarket Storage\'s own market label exactly; omit if unsure.' },
+        cerca: { type: 'string', description: 'Full-text search across disclosure titles and attachment contents.' },
+        date_from: { type: 'string', description: 'Only disclosures published on or after this date (YYYY-MM-DD, inclusive). If date_to is omitted, this tool searches just this one day.' },
+        date_to: { type: 'string', description: 'Only disclosures published on or before this date (YYYY-MM-DD, INCLUSIVE — the last day you want included). The upstream source\'s own date_to parameter is EXCLUSIVE of that day, so this tool always adds one calendar day internally before querying; the response echoes both your inclusive date_to and the exclusive bound actually sent upstream.' },
+        limit: { type: ['number', 'string'], description: `Number of disclosures to return (1-${PAGE_SIZE}). Default 10. The underlying source pages in fixed blocks of ${PAGE_SIZE}; a limit/offset combination that crosses a page boundary returns only the remainder of that page — use has_more and increase offset by ${PAGE_SIZE} to page further.` },
+        offset: { type: ['number', 'string'], description: 'Result offset for pagination. Default 0.' },
+      },
+    },
+  },
+  {
+    name: 'emarket_get_disclosure',
+    description:
+      'Fetch a single eMarket Storage disclosure by its Teleborsa protocollo id and publication date, as returned by emarket_search_disclosures. This source has no by-id lookup endpoint, only by-day listings, so both id and date are required. Returns the shaped record — company, title, category (if known), timestamp, and the document (PDF) URL.',
+    inputSchema: {
+      type: 'object',
+      properties: {
+        id: { type: 'string', description: 'Teleborsa protocollo id, e.g. "189482", from an emarket_search_disclosures result.' },
+        date: { type: 'string', description: "The disclosure's publication date (YYYY-MM-DD), from the same search result (published_at_local's date part)." },
+      },
+      required: ['id', 'date'],
+    },
+  },
+];
+
+async function callTool(name: string, args: Record<string, unknown>): Promise<unknown> {
+  try {
+    switch (name) {
+      case 'emarket_search_disclosures':
+        return await searchDisclosures(args);
+      case 'emarket_get_disclosure':
+        return await getDisclosure(args);
+      default:
+        return { error: `Unknown tool: ${name}` };
+    }
+  } catch (e) {
+    return { error: e instanceof Error ? e.message : String(e) };
+  }
+}
+
+// ---- issuer (azienda) picklist -- name/id map ------------------------------
+
+interface Issuer { id: string; name: string }
+
+let aziendaCache: { list: Issuer[]; until: number } | null = null;
+const AZIENDA_TTL_MS = 24 * 60 * 60 * 1000; // the issuer roster changes rarely; a day-old cache is fine for name resolution.
+
+async function getAziendaMap(): Promise<Issuer[]> {
+  if (aziendaCache && aziendaCache.until > Date.now()) return aziendaCache.list;
+  const res = await pwFetch(`${ORIGIN}${LISTING_PATH}`);
+  if (!res.ok) throw await httpError(res, 'eMarket Storage (Teleborsa OAM)');
+  const html = await res.text();
+  const list = parseAziendaSelect(html);
+  aziendaCache = { list, until: Date.now() + AZIENDA_TTL_MS };
+  return list;
+}
+
+function parseAziendaSelect(html: string): Issuer[] {
+  const selectMatch = /<select[^>]*id="edit-azienda"[^>]*>([\s\S]*?)<\/select>/.exec(html);
+  if (!selectMatch) return [];
+  const list: Issuer[] = [];
+  const optRe = /<option value="(\d+)">([^<]*)<\/option>/g;
+  let m: RegExpExecArray | null;
+  while ((m = optRe.exec(selectMatch[1]))) {
+    const nm = stripMarkup(m[2]).trim();
+    if (nm) list.push({ id: m[1], name: nm });
+  }
+  return list;
+}
+
+// ---- listing parsing --------------------------------------------------------
+
+interface RawRow {
+  protocollo: string;
+  aziendaId: string | null;
+  company: string;
+  title: string;
+  documentUrl: string | null;
+  publishedAtLocal: string | null;
+}
+
+function extractRows(html: string): RawRow[] {
+  const marker = 'class="views-row"';
+  const rows: RawRow[] = [];
+  let idx = html.indexOf(marker);
+  while (idx !== -1) {
+    const nextIdx = html.indexOf(marker, idx + marker.length);
+    let end = nextIdx;
+    if (end === -1) {
+      const pagerIdx = html.indexOf('<nav class="pager"', idx);
+      end = pagerIdx === -1 ? html.length : pagerIdx;
+    }
+    const block = html.slice(idx, end);
+    const row = parseRow(block);
+    if (row) rows.push(row);
+    idx = nextIdx;
+  }
+  return rows;
+}
+
+function parseRow(block: string): RawRow | null {
+  const protoM = /data-protocollo="(\d+)"/.exec(block);
+  if (!protoM) return null;
+  const aziendaM = /comunicati-finanziari\?azienda=(\d+)/.exec(block);
+  const pdfM = /href="(\/sites\/default\/files\/comunicati\/[^"]+\.pdf)"/.exec(block);
+  const timeM = /<time[^>]*>([^<]+)<\/time>/.exec(block);
+  const companyM = /class="news-azienda"><a[^>]*>([^<]*)<\/a>/.exec(block);
+  const titleM = /class="news-title"><a[^>]*>([^<]*)<\/a>/.exec(block);
+  return {
+    protocollo: protoM[1],
+    aziendaId: aziendaM ? aziendaM[1] : null,
+    company: companyM ? stripMarkup(companyM[1]).trim() : '',
+    title: titleM ? stripMarkup(titleM[1]).trim() : '',
+    documentUrl: pdfM ? `${ORIGIN}${pdfM[1]}` : null,
+    publishedAtLocal: timeM ? isoFromItalianDateTime(timeM[1].trim()) : null,
+  };
+}
+
+// The listing's visible timestamp is "DD/MM/YYYY - HH:MM" in Italian local
+// time (Europe/Rome). The row's `datetime` attribute is NOT usable — it is
+// truncated to something like "34Z" on every row observed (an upstream
+// templating bug), so this parses the human-readable text instead.
+function isoFromItalianDateTime(s: string): string | null {
+  const m = /^(\d{2})\/(\d{2})\/(\d{4})\s*-\s*(\d{2}):(\d{2})$/.exec(s);
+  if (!m) return null;
+  const [, dd, mm, yyyy, HH, MM] = m;
+  return `${yyyy}-${mm}-${dd}T${HH}:${MM}:00`;
+}
+
+function parseListing(html: string): { rows: RawRow[]; hasNext: boolean; lastPageIndex: number | null } {
+  const rows = extractRows(html);
+  const hasNext = /pager__item--next/.test(html);
+  const lastM = /pager__item--last[\s\S]{0,400}?page=(\d+)/.exec(html);
+  return { rows, hasNext, lastPageIndex: lastM ? Number(lastM[1]) : null };
+}
+
+async function fetchListing(params: URLSearchParams): Promise<{ rows: RawRow[]; hasNext: boolean; lastPageIndex: number | null }> {
+  const res = await pwFetch(`${ORIGIN}${LISTING_PATH}?${params.toString()}`);
+  if (!res.ok) throw await httpError(res, 'eMarket Storage (Teleborsa OAM)');
+  return parseListing(await res.text());
+}
+
+// Add one calendar day to a YYYY-MM-DD string (UTC arithmetic — the source's
+// date filter operates on calendar dates, not instants, so timezone drift
+// here does not matter).
+function addOneDay(dateStr: string): string {
+  const [y, m, d] = dateStr.split('-').map(Number);
+  const dt = new Date(Date.UTC(y, m - 1, d));
+  dt.setUTCDate(dt.getUTCDate() + 1);
+  return dt.toISOString().slice(0, 10);
+}
+
+function shapeRow(r: RawRow, categoryArg: string | undefined): Record<string, unknown> {
+  return {
+    id: r.protocollo,
+    company: r.company,
+    azienda_id: r.aziendaId,
+    title: r.title,
+    // The listing does not expose a per-row category — only the search
+    // filter does. When a category filter was applied, every returned row
+    // genuinely belongs to it (that is what the filter guarantees), so it is
+    // safe to echo; on an unfiltered search this is honestly null rather than
+    // guessed.
+    category: categoryArg ?? null,
+    category_label: categoryArg ? (CATEGORY_LABELS[categoryArg] ?? null) : null,
+    published_at_local: r.publishedAtLocal,
+    timezone: 'Europe/Rome',
+    document_url: r.documentUrl,
+  };
+}
+
+function strArg(v: unknown): string | undefined {
+  if (typeof v === 'string') {
+    const t = v.trim();
+    return t ? t : undefined;
+  }
+  if (typeof v === 'number' && Number.isFinite(v)) return String(v);
+  return undefined;
+}
+
+function clampInt(v: unknown, dflt: number, min: number, max: number): number {
+  let n: number;
+  if (typeof v === 'number' && Number.isFinite(v)) n = Math.trunc(v);
+  else if (typeof v === 'string' && v.trim() && Number.isFinite(Number(v))) n = Math.trunc(Number(v));
+  else return dflt;
+  return Math.min(max, Math.max(min, n));
+}
+
+async function searchDisclosures(args: Record<string, unknown>): Promise<unknown> {
+  const companyQuery = strArg(args.company);
+  const aziendaIdArg = strArg(args.azienda_id);
+  const categoryArg = strArg(args.category);
+  const marketArg = strArg(args.market);
+  const cercaArg = strArg(args.cerca);
+  const dateFrom = strArg(args.date_from);
+  const dateToArg = strArg(args.date_to);
+  const limit = clampInt(args.limit, 10, 1, PAGE_SIZE);
+  const offset = clampInt(args.offset, 0, 0, 100000);
+
+  // date_to in THIS tool's contract is INCLUSIVE (the last day to include),
+  // but the upstream `data_to` parameter is EXCLUSIVE of the day itself —
+  // and that mismatch is a genuine SILENT-LOSS bug, not just the single-day
+  // case: verified live 2026-09-24, date_from=2026-09-21&date_to=2026-09-22
+  // (a caller asking for "the 21st through the 22nd") returned 13 rows, ALL
+  // from the 21st — sending date_to=2026-09-22 upstream silently dropped the
+  // entire 22nd. Always send upstream a bound one calendar day PAST the
+  // caller's inclusive date_to; a lone date_from with no date_to means "just
+  // that one day" (inclusiveDateTo defaults to dateFrom in that case).
+  const inclusiveDateTo = dateToArg ?? dateFrom;
+  const upstreamDateTo = inclusiveDateTo ? addOneDay(inclusiveDateTo) : undefined;
+
+  let aziendaId: string | undefined = aziendaIdArg;
+  let resolvedCompany: string | undefined;
+  let candidates: { azienda_id: string; company: string }[] | undefined;
+
+  if (!aziendaId && companyQuery) {
+    const issuers = await getAziendaMap();
+    const ranked = rankMatches(companyQuery, issuers, (e) => e.name);
+    if (ranked.length === 0) {
+      return {
+        error: `"${companyQuery}" does not match any eMarket Storage issuer.`,
+        note: 'This pack only covers issuers that disclose through eMarket Storage. A separate mechanism, 1INFO, covers a different set of Italian issuers (e.g. IREN, SOMEC) and is not included here — the company may be a 1INFO issuer instead.',
+      };
+    }
+    const top = ranked[0];
+    const tiedAtTop = ranked.filter((r) => r.score === top.score);
+    if (tiedAtTop.length > 1) {
+      return {
+        error: `"${companyQuery}" matches more than one eMarket Storage issuer equally well.`,
+        candidates: tiedAtTop.slice(0, 10).map((r) => ({ azienda_id: r.item.id, company: r.item.name })),
+      };
+    }
+    aziendaId = top.item.id;
+    resolvedCompany = top.item.name;
+    candidates = ranked.slice(0, 5).map((r) => ({ azienda_id: r.item.id, company: r.item.name }));
+  }
+
+  const page = Math.floor(offset / PAGE_SIZE);
+  const withinPageOffset = offset - page * PAGE_SIZE;
+
+  const params = new URLSearchParams();
+  if (dateFrom) params.set('data_from', dateFrom);
+  if (upstreamDateTo) params.set('data_to', upstreamDateTo);
+  if (aziendaId) params.set('azienda', aziendaId);
+  if (categoryArg) params.set('categoria', categoryArg);
+  if (marketArg) params.set('mercato', marketArg);
+  if (cercaArg) {
+    params.set('cerca', cercaArg);
+    params.set('tipo_ricerca', 'fulltext');
+  }
+  if (page > 0) params.set('page', String(page));
+
+  const { rows, hasNext, lastPageIndex } = await fetchListing(params);
+  const pageSlice = rows.slice(withinPageOffset, withinPageOffset + limit);
+  const shaped = pageSlice.map((r) => shapeRow(r, categoryArg));
+
+  return {
+    count: shaped.length,
+    limit,
+    offset,
+    has_more: hasNext || withinPageOffset + limit < rows.length,
+    page_size: PAGE_SIZE,
+    // A true LOWER BOUND, never a fabricated exact total: every page but the
+    // last is a full PAGE_SIZE (verified live), so `last_page_index` pages of
+    // PAGE_SIZE rows are guaranteed to exist even before counting the last one.
+    ...(lastPageIndex !== null ? { at_least_total: lastPageIndex * PAGE_SIZE } : {}),
+    ...(dateFrom ? { date_from: dateFrom } : {}),
+    // date_to echoed here is the caller's INCLUSIVE end date (defaulting to
+    // date_from when omitted); upstream_date_to_exclusive is the actual bound
+    // sent to the source (one day later), shown for anyone debugging against
+    // the raw upstream directly.
+    ...(inclusiveDateTo ? { date_to: inclusiveDateTo, upstream_date_to_exclusive: upstreamDateTo } : {}),
+    ...(aziendaId ? { azienda_id: aziendaId } : {}),
+    ...(resolvedCompany ? { company_resolved: resolvedCompany } : {}),
+    ...(candidates && candidates.length > 1 ? { other_candidates: candidates.filter((c) => c.azienda_id !== aziendaId) } : {}),
+    ...(categoryArg ? { category: categoryArg, category_label: CATEGORY_LABELS[categoryArg] ?? null } : {}),
+    disclosures: shaped,
+  };
+}
+
+async function getDisclosure(args: Record<string, unknown>): Promise<unknown> {
+  const id = strArg(args.id);
+  const date = strArg(args.date);
+  if (!id) throw new Error('emarket_get_disclosure requires "id" — a Teleborsa protocollo id like "189482", from emarket_search_disclosures results.');
+  if (!date) throw new Error('emarket_get_disclosure requires "date" — the disclosure\'s publication date (YYYY-MM-DD). This source has no by-id lookup, only by-day listings.');
+
+  const dateTo = addOneDay(date);
+  const MAX_PAGES = 15; // bounds the worst case (a very busy day for the whole market) to ~15 sequential fetches.
+  for (let page = 0; page < MAX_PAGES; page++) {
+    const params = new URLSearchParams({ data_from: date, data_to: dateTo });
+    if (page > 0) params.set('page', String(page));
+    const { rows, hasNext } = await fetchListing(params);
+    const hit = rows.find((r) => r.protocollo === id);
+    if (hit) return shapeRow(hit, undefined);
+    if (!hasNext) break;
+  }
+  return { error: 'disclosure not found for that id/date', id, date };
+}
+
+export default { tools, callTool, meter: { credits: 1 } } satisfies McpToolExport;