Global Trade Alert

Discover published GTA reports/blog posts without reading their prose. Use this to resolve a named or described publication to one or more exact ``report_id`` values; the chosen IDs are what ``gta_search_analyses`` takes. Ambiguous candidates are returned as a list, not resolved to one. For live policy records or counts, use the intervention tools instead.

gta_find_analyses

Fetch complete details for one or more GTA interventions by ID. Returned records use their semantic ``intervention_id`` as the linked handle in the workbench citation system. Pass `intervention_id` for a single record or `intervention_ids` for an ordered batch with partial-success errors. Exactly one ID argument is required. Use `show_keys` to project fields and control response width. Markdown always retains the ID, title, and announcement date required for readable citations; JSON preserves the requested projection plus mandatory provenance. Available fields include intervention_id, state_act_id, state_act_title, intervention_type, gta_evaluation, mast_chapter, dates, jurisdictions, products, sectors, description, sources, and canonical URLs.

gta_get_intervention

Extract granular impact chains showing implementing-product/sector-affected jurisdiction tuples. Unlike the main data endpoint which aggregates data, impact chains provide unaggregated relationships between implementing jurisdictions, affected products/sectors, and affected jurisdictions. Essential for bilateral trade flow analysis and detailed impact assessment.

gta_get_impact_chains

Fetch the complete description-update history for a specific GTA intervention. The response ends with the semantic ``intervention_id`` as the linked handle for the workbench citation system. The GTA database splits each intervention's description into ordered, timestamped updates (status: new / changed / deleted / static). Use this tool when you need to see HOW an intervention evolved over time — e.g. audit trails, policy tracking, change analysis, or understanding what was added when analysts revised the record. For the current combined description, use `gta_get_intervention`. For a global feed of recent updates across all interventions, use `gta_list_ticker_updates`.

gta_get_intervention_updates

Use AI to identify HS product codes from a natural language product description. Unlike `gta_lookup_hs_codes` (keyword matching), this tool uses semantic understanding to map descriptions like "electric vehicles" or "lithium-ion batteries" to the correct HS codes — even when the official HS terminology differs from common language. Use this tool when: - The user describes products in plain language and `gta_lookup_hs_codes` returns no/poor results - You need to extract HS codes from a policy document excerpt - The product doesn't have an obvious keyword match in the HS nomenclature Use `gta_lookup_hs_codes` instead when: - You already know the keyword (e.g., 'lithium', 'steel') - You want to browse a specific HS chapter by number

gta_guess_hs_codes

Return a structured query PLAN for a GTA-related natural-language question WITHOUT calling the API. Surfaces the filters the user's question implies, the defaults that would apply, and disambiguation questions worth answering before executing a search. Applies when the user's question contains explicit ambiguity terms along the GTA taxonomy; the plan lists the filters `gta_search_interventions` and `gta_count_interventions` would take. Trigger phrases: - Targeted vs affecting: "targeting", "against", "aimed at", "affecting", "impacting" - Harmful vs liberalising: "harmful", "protectionism", "trade barrier", "discrimination", "liberalising", "openness", "market opening" - Lifecycle: "active", "in force", "currently", "proposed", "pending", "revoked", "withdrawn", "terminated" - Trade flow: "export ban / restriction / control / tax / quota", "export subsidy / support / credit / financing", "import tariff / duty / restriction / quota" - Trade defence: "anti-dumping", "safeguard", "countervailing", "anti-subsidy" - Counting unit: "how many announcements", "how many state acts", "how many decrees" Not for a question that is already concrete (e.g. "show me US tariffs on Chinese EVs from 2024"): `gta_search_interventions` takes that directly. Returns markdown by default. JSON returns the structured filter dict suitable for direct passing to gta_search_interventions or gta_count_interventions.

gta_interpret_query

List the proposed / parked tariff actions a client can activate per request. - Scope: US imports from 2025-01-20 only; one rate `tariff_calculate`, over time `tariff_timeline`, averages `tariff_aggregate` or `tariff_world_timeline` (reference: us-tariff://guide/common-queries). - What-ifs: a deal what-if is `tariff_create_scenario`, `tariff_toggle_deal`, then a read tool with `scenario_id`; the baseline is read-only; a proposed measure is no deal: pass its `tariff_list_proposals` name to `activate_proposals` (reference: us-tariff://guide/scenario-workflow). - Quoted figures: a quoted report figure is context; the rate comes from a tool call (reference: us-tariff://guide/evidence-vs-context). - Traps: India is 699, not 356; a Section 301 of 0% is carved out or not yet in force, never absent; a banned Canadian line reads 0% (reference: us-tariff://reference/critical-gotchas). Applies when the user asks about a *proposed*, *announced*, or *not-yet-in-force* measure — e.g. "what would Brazilian exports face if the announced Section 301 enters into force", "model the proposed forced-labour tariffs". It returns the activatable actions (currently `bra_s301`, `fl_s301`) with a description of each; the name is what `activate_proposals` takes on `tariff_calculate`, `tariff_aggregate`, or `tariff_country_brief`. `tariff_toggle_deal` toggles bilateral deals (e.g. `brazil_topup`, the legacy IEEPA top-up, NOT the Section 301), so it does not model these measures, and a hand-computed trade-weighted rate is not the engine's. The engine derives the authoritative activated rate from the activation path. Parked measures (reference: us-tariff://reference/proposed-and-parked-measures).

tariff_list_proposals

Discover GTA editorial threads and curated frameworks for analyst concepts. Threads are maintained editorial cohorts: the curated definition of an analyst concept such as government guidance funds, sanctions, or a policy initiative, which a free-text query only approximates. The returned thread ID is what the `threads` filter of GTA search/count takes.

gta_list_threads

Monitor which GTA interventions were recently modified, via the ticker endpoint. Returned records use their semantic ``intervention_id`` as the linked handle in the workbench citation system. This is a lightweight triage feed: each row reports WHICH intervention changed, WHEN (modified date), and the update STATUS (new / changed / static) — it does NOT return the changed description text. This keeps the listing pass off the usage meter. To read the actual text of an update, drill into a specific intervention with `gta_get_intervention_updates`. Useful for monitoring policy evolution and spotting which measures to inspect next. Each row is one description update, so a record can repeat. `update_status` describes that update, not the policy: a 2010 record can carry a `new` row. `modified` dates the record update; `date_announced` dates the policy.

gta_list_ticker_updates

List the full catalogue of readable GTA and US-tariff documents — every static resource and URI template on this server, with its name and description: the query guides and each family's full guide index, the reference tables (jurisdictions, sectors, MAST chapters, intervention types, country codes) and the methodology explainers. Any URI listed here is a valid `uri` for `read_resource`.

list_resources

Search CPC (Central Product Classification) sector codes by keyword or code prefix. Use when the user asks about services or broad sector categories; returns the group IDs the `affected_sectors` filter of gta_search_interventions takes. Divisions >= 50 and groups >= 500 are services; the rest are goods. Examples: - search_term='financial' → finds CPC 711, 715, 717 - search_term='71' → lists all sectors in division 71 (financial services) - search_term='transport' → finds transport-related CPC sectors Returns a markdown table with codes and a ready-to-use affected_sectors list of group IDs (use_with); a division matched without groups expands to all its groups.

gta_lookup_sectors

Search HS (Harmonized System) product codes by keyword, chapter number, or code prefix. Use when the user names specific commodities or products; returns matching codes across all 3 levels (chapters, headings, subheadings) with the numeric IDs the `affected_products` filter of gta_search_interventions takes. If this tool returns no results or irrelevant matches for a natural language product description (e.g., 'electric vehicles', 'green technology'), use `gta_guess_hs_codes` instead — it uses AI semantic matching to find codes even when official HS terminology differs from everyday language. Examples: - query='lithium' → finds HS 282520, 283691, etc. - query='28' → lists all codes in chapter 28 (inorganic chemicals) - query='8541' → lists subheadings under heading 8541 (semiconductors) - query='steel' → finds relevant iron/steel HS codes Returns a markdown table with codes and a ready-to-use affected_products list.

gta_lookup_hs_codes

List US HTS 8-digit products whose code starts with `search`. Pass `searches` to resolve several independent code prefixes in one call. Each prefix keeps the same matching and per-search `limit` behavior. USE THIS TOOL when you need to: - list all HS8 within a chapter (`search="87"`) or heading (`search="8703"`), - confirm an HS8 exists for `tariff_calculate`. Descriptions come back as placeholder strings (`"HS {code}"`) until the backend enriches them. The HS floor is 8 digits; 10-digit HTS is truncated upstream. Method (reference: us-tariff://reference/methodology).

tariff_lookup_products

Resolve a natural-language phrase to one or more GTA intervention type names + IDs. Use when the user asks about a category of measures and the exact GTA term is unknown: the `intervention_types` filter of `gta_search_interventions` and `gta_count_interventions` takes specific type names ("Anti-dumping", "Financial grant", "Import tariff") or numeric IDs, not free-text. The GTA type vocabulary is foreign to most users; this tool bridges the gap. The search is fuzzy: it matches on full substring, on word tokens, and (lower priority) on partial token overlap. Returns matches ordered by relevance.

gta_lookup_intervention_types

Resolve a country name, ISO3 code, or UN M49 code; show deal flags. USE THIS TOOL when you need to: - confirm the UN M49 code the tariff engine wants for a country, - check which bilateral deals affect a given country, - list countries matching a fragment ("list south american countries that have a deal"). **Gotcha:** GTA's UN codes are NOT always the standard UN M49 value users expect. The authoritative code is whatever `/reference/countries/` returns — this tool surfaces that. **India = 699 in GTA, not 356.** This tool is the source for the UN values the other `tariff_*` calls take. **Deal flags shown:** UK, JPN, EU, KOR, CHE, BRA, USMCA (reference: us-tariff://reference/bilateral-deals). UN, ISO3 and GTA codes diverge for a few countries (reference: us-tariff://reference/country-codes).

tariff_lookup_countries

Read one GTA or US-tariff reference document by URI. The query guides, by topic and rule: GTA intervention records, incl. announced tariffs: - policy types and countries: subsidies → `mast_chapters=['L']`, export controls → `['P']`, tariffs → `intervention_types=['Import tariff']`, harmful or 'imposed' barriers → `gta_evaluation=['Harmful']`, actor → `implementing_jurisdictions`, target → `affected_jurisdictions` (reference: gta://guide/query-intent-mapping) - products and services: services take CPC `affected_sectors` (ids 500+), never HS; specific goods take HS `affected_products`, broad categories CPC (reference: gta://guide/cpc-vs-hs) - exclusions: list the values, set the matching keep_* to false — keep_* parameters, not a negated query string (reference: gta://guide/exclusion-filters) - names: a name no filter carries goes in `query` as an exact phrase; & or | loosens it (reference: gta://guide/query-syntax) - dates and trends: new measures and trends: `date_announced_*`; took effect in a period: `date_implemented_*` (reference: gta://guide/date-fields) US applied tariff rates, 2025-01-20 on: - scope and tool choice: US imports from 2025-01-20 only; one rate `tariff_calculate`, over time `tariff_timeline`, averages `tariff_aggregate` or `tariff_world_timeline` (reference: us-tariff://guide/common-queries) - what-ifs: a deal what-if is `tariff_create_scenario`, `tariff_toggle_deal`, then a read tool with `scenario_id`; the baseline is read-only; a proposed measure is no deal: pass its `tariff_list_proposals` name to `activate_proposals` (reference: us-tariff://guide/scenario-workflow) - quoted report figures: a quoted report figure is context; the rate comes from a tool call (reference: us-tariff://guide/evidence-vs-context) - known traps: India is 699, not 356; a Section 301 of 0% is carved out or not yet in force, never absent; a banned Canadian line reads 0% (reference: us-tariff://reference/critical-gotchas) `list_resources` lists every guide, reference table and explainer.

read_resource

Suggest relevant GTA website pages from the site's approved route catalogue. This does not invent URLs. Use it only when the user enabled GTA page suggestions; filtered Data Centre/Activity Tracker links remain the preferred way to reproduce an actual calculation or inspect its records.

gta_recommend_pages

Retrieve grounded passages from one or more published GTA analyses. This tool reports what dated GTA publications found, argued, or concluded; it is not a source for live intervention totals. Named publications resolve to report IDs through ``gta_find_analyses``. Passages carry the report/blog/PDF URLs and publication dates. Unknown IDs fail closed rather than widening the search.

gta_search_analyses

Search GTA INTERVENTION records; ``intervention_id`` is the workbench citation system handle. Aggregates: `gta_count_interventions`. DO NOT USE for US applied tariff rates from 2025-01-20 onwards -- use `tariff_*`. Fine for a non-US importer, an earlier date, or announcements. Filters: - Policy types: subsidies → `mast_chapters=['L']`, export controls → `['P']`, tariffs → `intervention_types=['Import tariff']`, harmful or 'imposed' barriers → `gta_evaluation=['Harmful']`, actor → `implementing_jurisdictions`, target → `affected_jurisdictions` (reference: gta://guide/query-intent-mapping). - Products: services take CPC `affected_sectors` (ids 500+), never HS; specific goods take HS `affected_products`, broad categories CPC (reference: gta://guide/cpc-vs-hs). - Exclusions: list the values, set the matching keep_* to false — keep_* parameters, not a negated query string (reference: gta://guide/exclusion-filters). - Names: a name no filter carries goes in `query` as an exact phrase; & or | loosens it (reference: gta://guide/query-syntax). - Dates: new measures and trends: `date_announced_*`; took effect in a period: `date_implemented_*` (reference: gta://guide/date-fields). TIMELINE: the description is ordered updates; a new event may update an old record: check coverage without date filters; `latest_action` flags it. Read all: `gta_get_intervention`. gta_evaluation: 'Red'/'Amber' harmful, 'Green' liberalising. Every parameter, including status and affected_flow (reference: gta://guide/parameters). COMPLETENESS: a state act can hold siblings under other classifications; an HS lookup does not prove derivatives share the parent code. Package/product questions need a second pass (reference: gta://guide/common-mistakes). ZERO RESULTS: pre-2008 dates, trade agreements, sub-threshold subsidies, non-discriminatory measures, wrong taxonomy or publication lag can empty a live corpus (reference: gta://reference/database-scope). Only `absence_within_scope` licenses an absence claim.

gta_search_interventions

Semantic vector search over GTA intervention descriptions. Returned records use their semantic ``intervention_id`` as the linked handle in the workbench citation system. Ranks interventions by how well their FULL TEXT matches a natural-language query. Use this for thematic or conceptual searches that structured filters cannot capture (e.g. "supply chain resilience", "green hydrogen incentives", "EV battery manufacturing"). Combines with gta_search_interventions or gta_count_interventions: structured filters (country, date, MAST chapter, etc.) narrow the candidate pool, and the intervention IDs they return, passed here via intervention_ids, are ranked by thematic relevance.

gta_semantic_search

Identify which parameters and regimes move the tariff rate most. - Scope: US imports from 2025-01-20 only; one rate `tariff_calculate`, over time `tariff_timeline`, averages `tariff_aggregate` or `tariff_world_timeline` (reference: us-tariff://guide/common-queries). - What-ifs: a deal what-if is `tariff_create_scenario`, `tariff_toggle_deal`, then a read tool with `scenario_id`; the baseline is read-only; a proposed measure is no deal: pass its `tariff_list_proposals` name to `activate_proposals` (reference: us-tariff://guide/scenario-workflow). - Quoted figures: a quoted report figure is context; the rate comes from a tool call (reference: us-tariff://guide/evidence-vs-context). - Traps: India is 699, not 356; a Section 301 of 0% is carved out or not yet in force, never absent; a banned Canadian line reads 0% (reference: us-tariff://reference/critical-gotchas). USE THIS TOOL when the user asks "which policy matters most for tariff on X from Y", "what drives this rate", "if one thing changed, which would move the tariff most". For several products from the same country, pass `hs_codes`. Each product gets its own ranking with the same `top_n` limit. **How it works.** Sweeps each tunable parameter in the baseline kit to its min/max and records the rate delta; for each deal in force, reports its `counterfactual_final_rate`. Ranks all scenarios by absolute delta, top N. Read-only -- no scenarios are created. **Interpretation note.** A negative delta means the driver currently **reduces** the rate; positive means it **adds**. Quoting: the rate is a modelled applied rate under the regimes in force on `policy_date`, not the statutory MFN rate and not a prediction (reference: us-tariff://guide/interpretation-rules). Parameters (reference: us-tariff://reference/tunable-parameters); deals (reference: us-tariff://reference/bilateral-deals).

tariff_sensitivity

Show how applied rates changed over time by product for one country. - Scope: US imports from 2025-01-20 only; one rate `tariff_calculate`, over time `tariff_timeline`, averages `tariff_aggregate` or `tariff_world_timeline` (reference: us-tariff://guide/common-queries). - What-ifs: a deal what-if is `tariff_create_scenario`, `tariff_toggle_deal`, then a read tool with `scenario_id`; the baseline is read-only; a proposed measure is no deal: pass its `tariff_list_proposals` name to `activate_proposals` (reference: us-tariff://guide/scenario-workflow). - Quoted figures: a quoted report figure is context; the rate comes from a tool call (reference: us-tariff://guide/evidence-vs-context). - Traps: India is 699, not 356; a Section 301 of 0% is carved out or not yet in force, never absent; a banned Canadian line reads 0% (reference: us-tariff://reference/critical-gotchas). FIRST-CHOICE TOOL for any US applied-tariff question post-2025-01-20. USE THIS TOOL when the user asks "how has the tariff on X from Y changed", "trace the tariff over 2025 for X from Y", "what was the rate on date Z vs today". For several products from the same country, pass `hs_codes`. Each product gets its own timeline over the same selected policy dates. DO NOT use for: - single-date queries (use `tariff_calculate`) - multi-country comparisons (use `tariff_compare`) - **world / country / chapter *average* over time — use `tariff_world_timeline`**. This tool loops per-product per-date and is slow/wrong for aggregates. **Behavior.** Gets the authoritative regime-activation dates, then calls the single-kit endpoint at each date in range. Returns a table of rate moves per regime change, with the formula identifier per row. Regimes (reference: us-tariff://reference/policy-regimes); formulas (reference: us-tariff://reference/rate-formulas). Quoting: the rate is a modelled applied rate under the regimes in force on `policy_date`, not the statutory MFN rate and not a prediction (reference: us-tariff://guide/interpretation-rules).

tariff_timeline

Enable or disable a bilateral deal for a custom tariff scenario. - Scope: US imports from 2025-01-20 only; one rate `tariff_calculate`, over time `tariff_timeline`, averages `tariff_aggregate` or `tariff_world_timeline` (reference: us-tariff://guide/common-queries). - What-ifs: a deal what-if is `tariff_create_scenario`, `tariff_toggle_deal`, then a read tool with `scenario_id`; the baseline is read-only; a proposed measure is no deal: pass its `tariff_list_proposals` name to `activate_proposals` (reference: us-tariff://guide/scenario-workflow). - Quoted figures: a quoted report figure is context; the rate comes from a tool call (reference: us-tariff://guide/evidence-vs-context). - Traps: India is 699, not 356; a Section 301 of 0% is carved out or not yet in force, never absent; a banned Canadian line reads 0% (reference: us-tariff://reference/critical-gotchas). USE THIS TOOL to remove or reinstate a deal in your scenario — useful for counterfactuals like "what if the UK deal had never been signed" or "what if the EU deal were extended to Switzerland". Valid deal names are constrained by the `DealName` literal (reference: us-tariff://reference/bilateral-deals). **Auth requirement.** Write-capable API key or `mcp:write` OAuth scope. The baseline (`id=1`) is read-only — trying to toggle on it returns 403. Methodology (reference: us-tariff://explain/methodology-overview); assumptions (reference: us-tariff://explain/assumptions-register).

tariff_toggle_deal

World, country, or chapter trade-weighted tariff timeline — ONE round-trip. - Scope: US imports from 2025-01-20 only; one rate `tariff_calculate`, over time `tariff_timeline`, averages `tariff_aggregate` or `tariff_world_timeline` (reference: us-tariff://guide/common-queries). - What-ifs: a deal what-if is `tariff_create_scenario`, `tariff_toggle_deal`, then a read tool with `scenario_id`; the baseline is read-only; a proposed measure is no deal: pass its `tariff_list_proposals` name to `activate_proposals` (reference: us-tariff://guide/scenario-workflow). - Quoted figures: a quoted report figure is context; the rate comes from a tool call (reference: us-tariff://guide/evidence-vs-context). - Traps: India is 699, not 356; a Section 301 of 0% is carved out or not yet in force, never absent; a banned Canadian line reads 0% (reference: us-tariff://reference/critical-gotchas). FIRST-CHOICE TOOL for any US applied-tariff question post-2025-01-20. USE THIS TOOL when the user asks: - "draw me a timeline of the US average import tariff from X to Y" - "how has the average tariff on Germany evolved over 2025" - "show the chapter-72 (steel) average over the last year" DO NOT use: - `tariff_timeline` (per product × per country — wrong granularity; slow for aggregates) - `tariff_aggregate` (single date only — would require N loops) **Behaviour.** Calls `/api/v1/tau/aggregate/timeline/` which returns one row per policy-change date between `start_date` and `end_date`, each with the trade-weighted average final rate and its component breakdown. **Precision tiers.** `sample_size` trades speed for coverage: top-1000 flows cover ~60% of trade by value; bump to 5000 (the backend maximum) for policy-brief-grade accuracy. Full-universe averages run higher than sampled.

tariff_world_timeline