Connected research workflows

Open Research tools from My Watchtower or a workspace toolbar. The seven views connect existing monitoring, cases and public evidence. Normal account access and source entitlements apply. Switching accounts or entering replay clears the open research context and cancels pending requests.

Saved searches

1. In Pro Brief or Daily Brief, apply your keyword, topic, source-type and editorial choices. Click Save this search as a watch.
2. Review the exact saved filters, publication window and inherited publisher regions. Give the watch a name.
3. Choose next eligible in-app evaluation or a daily inbox digest. Explicitly enable notifications if wanted, then save. Open My Watchtower → Inbox for delivery.

Keywords use case-insensitive Unicode-normalized literal matching: all terms must match, quotes retain phrases, and -term or -"phrase" excludes them. The limit is 180 characters and 12 terms. Country/company words are contextual matches, not verified locations or holdings. A newsroom search cannot silently become a general watchlist with different matching rules.

The first healthy server evaluation establishes a baseline without alerting on old headlines. Canonical repeats and mirrored headlines are grouped; substantive revisions remain labelled. Monitoring checks at most 500 retained stories per search. Pro uses the rolling last 48 hours; Daily uses the current UTC day. The briefing story-count preference does not cap the monitoring scan.

Immediate means the next eligible server evaluation, not external push. Daily digests use your existing briefing IANA time zone and local hour, with up to 20 queued records per edition. A 15–1440 minute cooldown, per-watch budget and shared 10-delivery/day account budget apply. Queues are bounded to 100 matches and the account-private storage budget, and expire after 48 hours; omitted/expired counts are visible. Source delays and polling/load can delay delivery. This feature does not subscribe you to email or SMS.

Investigation dossiers

Dossiers are the existing War Room cases. Open one or create a new case, write a research question, and add assessment, notes and unanswered questions. Pin a reviewed dataset page, forecast scorecard or scenario relationship. In a workspace, Pin selection opens the applied public source in the coverage inspector; review it before pinning. To retain an individual article, inspect its /items/0 or other actual source pointer instead of assuming the first field page contains the article.

Pins preserve the source endpoint, applied parameters, source snapshot, original clocks when supplied, captured fields, analyst context and pin time. Pin time is not publication or event time. Each pin is limited to 48 KiB; choose a narrower pointer or page for larger data. Cases retain at most 200 evidence records and existing account-private storage limits. Evidence can support, contradict, contextualize or leave a question unresolved. Citation numbers are not reassigned after deletion.

Click Save dossier revision to retain edits across devices. Concurrent edits return a conflict instead of overwriting the saved case. Export an unsaved draft before reloading if you want to retain it. JSON and dated Markdown exports preserve the evidence and qualifications; they are analyst-authored products, not automatic verification.

After saving, share with a specific eligible account as a viewer or editor for 1, 7, 30 or 90 days. Copy the returned recipient-specific link. A link alone does not grant access to another account. Revoke shares from the same view. Deleting a case revokes its grants; account deletion uses the existing private-state barrier. Older clients that cannot preserve extended provenance cannot silently strip it.

Data, API and MCP coverage

Search Data coverage by workspace, dataset or topic. The catalogue is read from the currently reviewed MCP registry; workspace associations come from reviewed Assistant references. Missing workspace mappings and non-catalogued associations are shown explicitly. A reviewed route is not proof that all fields/history are available or current, and a source without a documented workspace mapping is not necessarily unused.

Select Inspect fields / history, review supported parameters, and read the live selection. The inspector provides equivalent native API and MCP requests. Returned JSON Pointers preserve original types, including zero, null and stale flags. Use an empty pointer for the root response. Follow Next field page or long-field continuation without changing the MCP snapshot. Native nextOffset, next_cursor and native source snapshot belong inside Parameters and paginate source records; they are independent of MCP field and text pagination. Read original source dates, coverage and units before drawing conclusions. A bounded inspected page is not a complete dataset download.

Account, credential, payment, private portfolio/case/report/settings, administrative, binary stream and mutation surfaces remain intentionally excluded from public MCP.

Applied research context

Apply country, explicit company symbol, topic, reference year and source-native horizon in Applied context, or capture a successfully applied workspace selection. The context strip makes the retained research focus visible across workspaces. Pending form edits and unavailable source requests are not accepted as applied data.

Send supported fields to Geopolitical Futures, Geopolitical Outlook, Public finance or Pro Brief news. Unsupported fields are reported instead of silently applied. Futures uses its supported next-year/2030/2040/2050 selector; Outlook uses 24h/7d/30d forecast horizons. Reference years and forecast horizons are different concepts. The company action opens the reviewed company evidence source for the explicit symbol. No name-to-ticker or parent/subsidiary inference is made.

Use context in Assistant opens a question containing this explicit context and the current workspace. Review Preview/source selections before Run. The existing account-bound AI connection is reused; opening the Assistant does not generate a paid answer or upload private dossier contents.

Forecast scorecards

Choose Futures, Forecast Lab funds, Systemic targets, Energy prices or Geopolitical Outlook. Read the exact target, original horizon unit, model version, historical evaluation windows, simple baselines, genuinely issued outcomes and calibration qualifications. Source-wide issuance counts are labelled separately from selected target/model partitions. Missing values remain unavailable, not zero.

Futures predicts source-coded event presence inside borders, not the aggressor or a specific attack. Fund targets retain their exchange-session and price-return definitions. Energy horizons count source observations, not calendar days; errors remain in native units. Authored geopolitical priors are not measured performance baselines. Historical current-vintage reconstructions are not prospective success, and source calibration gates remain authoritative. This view does not fit models, issue forecasts, resolve outcomes or rewrite any ledger.

Cross-device preference sync

In Device sync, choose scopes, preview them, then explicitly save them to your account or apply the saved version on another signed-in device. Scopes are Pro/Daily news choices, pinned workspaces/table preference, current device panel/tab layout and configurable Pro Brief window definitions, default display mode and briefing timezone/hour. Layout applies to the device's native layout. Existing watchlists, cases and briefing history use their own cloud stores and are not duplicated here.

The account document has an optimistic revision. If another device changes it, refresh, compare and choose again; no automatic last-writer winner is used. Applying cloud settings is also explicit and affects only selected scopes. Guest settings are not uploaded automatically. Browser storage failure is reported and prior values are restored where storage permits. Credentials, arbitrary local storage, recent browsing history, private cases, holdings and provider settings are excluded.

Scenario-to-exposure analysis

Choose a country, historical trade year, HS product chapter, importer/exporter role, explicit disruption percentage and resource reference year. Optionally select up to four supported company symbols, or copy explicit symbol subjects and a single country from a saved watch. Company-name phrases are not guessed into symbols.

The view presents up to 40 retained country/product relationships with original values, units, source year, relationship class and reported/estimated status. HS chapters describe products, not measured industry revenue exposure. Source models such as reconciled BACI trade remain estimates. The value-scaled hypothetical reference is simply source value × your assumed fraction: it is not a financial loss, observed affected volume, future flow or price forecast. Country-centre links are not physical shipping routes. Different relationship classes are not added.

Resource projections keep their original anchors and assumptions. Company legal headquarters, where documented, are separate from operating assets, suppliers, revenues and disruption exposure. Missing operational/project links remain unmapped. No holdings, parent/subsidiary exposures or undisclosed supplier networks are inferred. Pin relationships and assumptions to a dossier or export the complete bounded view.