GTA in Your AI Assistant
Generate trade & industrial policy insight based on verified evidence
Generate trade & industrial policy insight based on verified evidence
Most MCP queries cost nothing. Your allowance is drawn down only when you pull the full record of a specific intervention, not when you ask for statistics, monitoring, or search.
Free use: 0 records. Aggregate counts, monitoring feeds, search results, and product-code lookups never touch your allowance. In our own testing, 54% of MCP queries used no metered allowance at all. A typical free query: “How many harmful trade measures has the US announced per year since 2020?” The answer is a table by year, 1,576 in 2020 easing to 708 in 2025, built entirely from aggregate counts. Follow-ups that add a dimension (“break that down by measure type”, “compare that to China”) stay free too. Nothing is being read in full yet.
Light use: 1 to 5 records. Reading one intervention in full sits in this band. So does checking a product route against current restrictions. “Tell me the full details of India's latest solar panel import duty” costs one record: the search is free, and only the best match is read in full. Most compliance checks, such as “Can I export copper concentrate from Chile to China without restrictions?”, are answered by the free search summaries and cost nothing.
Heavy use: 4 to 18 records. Comparing policy across countries means reading interventions in full rather than skimming summaries. “Compare US and EU semiconductor subsidy programs” draws four to eight records: the search on both sides is free; only the measures that matter for the comparison are read in full. The most demanding case we have tested, a complete EV battery briefing across tariffs, subsidies, local-content rules, and export restrictions, drew 18 records.
Two rules hold in every band. First, re-reading is free: a follow-up on a record you already pulled does not draw on your allowance again. Second, sessions add up slowly. A broad scan, a sector cut, a country comparison, and a written summary typically total three to nine records across the whole conversation, not per question.
We want our data used. We waive the subscription fee for academic researchers and civil-society organisations who confirm their institutional affiliation, the same practice we use for research API keys. An academic email address or a letter on institutional letterhead is enough. For a civil-society organisation, a name and a mission we can verify independently.
If cost is the only thing standing between you and the data, write to us anyway. Tell us what you are working on and we will find a way to get you access.