Thirty-one open data sources, one reconciled dataset searchable by meaning, separate web research, and a method that shows its evidence and its limits. Written for readers who need to trust, reuse or check the results.
Every live source is read again when a brief is made, so a brief has each publisher's latest release on that day, and every card states when it was fetched. What differs is how often publishers release: some within hours, some once a year. Green dates are the newest periods in briefs made over the last 30 days (latest 9 Nov 2026); the newest period differs by country, so a country page can show an older one when its publisher has released nothing newer for it.
Update cycles are as each publisher states them; follow one to its statement.
When a standard source is behind, the brief adds a newer one beside it, never in place of it, and says which is which.
The core dataset is reconciled, stored and searched inside the platform, so figures are exact and every run is reproducible.
What: Every recorded agreement and disbursement, 2015–2025, with its partner, sector and recipient country.
How: Cleaned and checked in stages; the final totals match the published source exactly. Stored in the app's own database.
Updated: Refreshed after each update of the source.
Built from the same data
What: Every agreement's title and description, embedded with a multilingual text-embedding model, so Norwegian and English texts match.
How: Stored in the app's database. A brief searches it with the reader's purpose, restricted to the countries and years in scope.
Updated: Re-embedded when the data is refreshed.
OECD Data Explorer (SDMX API)
What: Each recipient country's full Creditor Reporting System series since 2015 (every donor and sector), donors' aid totals (DAC1), mobilised private finance and foundation projects, as the OECD's own answers.
How: Stored in the app's database with the date each was fetched; briefs filter donors, sectors and years from the stored copy, and use it when the OECD's server refuses a request.
Updated: Fetched again after 30 days (donors' totals after 7); a background job fills the remaining recipient countries a few each hour.
ForeignAssistance.gov complete dataset
What: Every US economic (non-military) aid transaction since fiscal year 2019: obligations and disbursements by country, sector, agency and implementing partner.
How: The full file is read into the app's database and summed by country, fiscal year, sector and partner; the file's own date is kept.
Updated: Checked daily; reloaded when the publisher's file changes (it publishes quarterly).
Written by the app
What: Watches, runs, every decision with its reason, briefs and feedback, and the briefs signed-in reviewers keep in Your briefs.
How: Stored in the app's database next to the data, so each run starts from the last one.
Updated: Written on every run; saved briefs are removed when deleted, after twelve months at the latest.
Each card in a brief carries its publisher, scope and retrieval date. A source that fails is shown as missing, never replaced.
| Source | What it measures | Scale | Updated | How the app gets it | Themes |
|---|---|---|---|---|---|
| OECD Creditor Reporting System OECDYearly The OECD's database of aid activities reported by its Development Assistance Committee members, multilateral organisations and some foundations. It is the standard reference for who gives aid to whom and for what, and most published aid statistics are built from it. Figures arrive about a year after the year ends. | Every official donor's disbursements by recipient and sector, and project-level records with channel and description | USD millions, current prices; the all-donor total also in constant USD | Yearly each December, for the preceding year | Each recipient's full series kept in the database and refreshed after 30 days; project records by live query, kept the same way | Funding |
| OECD DAC1 OECDYearly The OECD's headline aid totals for each donor, the source of the yearly figures on aid as a share of national income. It shows a donor's overall trend, not where the money goes. | Donors' total official development assistance, including the provisional latest year | USD millions | Yearly provisional totals each April, final figures each December | Kept in the database, refreshed weekly | Funding |
| US foreign assistance (ForeignAssistance.gov) US governmentQuarterly The US government's own record of its foreign assistance: every obligation and payment by country, sector, agency and implementing partner, published much sooner than the OECD's data. It shows the effect of the 2025 US cuts country by country; economic and military aid are kept apart. | US economic aid by fiscal year to the brief's countries: disbursed and newly obligated, in all sectors and in the brief's sectors, with the largest implementing partners; military aid excluded | USD millions, current; US fiscal years (October to September) | Quarterly agencies report obligations and disbursements each quarter (OMB Bulletin 12-01), and the app reloads the file when it changes | Loaded into the database from the complete file when it changes (checked daily) | Funding |
| OECD mobilised private finance OECDContinually The OECD's measure of private money brought in by public development finance, for example through guarantees or co-investment. It is the main reference for blended finance; amounts follow the OECD's attribution rules and are published here for groups of providers. | Private money mobilised by official development finance, by mechanism (guarantees, syndicated loans, credit lines, direct investment, fund shares, co-financing) and sector | USD millions; published for provider groups only | Continually providers report on staggered schedules | Live query per brief, kept in the database for 30 days | FundingEconomy and investment |
| OECD Private Philanthropy for Development OECDYearly Foundations that report their grants to the OECD in the same format as government donors, among them the largest ones. Useful for comparing peer foundations; most smaller foundations do not report. | Foundations that report to the OECD: their funding and projects in the brief's countries and sectors | USD millions; reporting foundations only (about 40+ of the largest), not all philanthropy | Yearly with the OECD's December development finance release | Live query per brief, kept in the database for 30 days | Funding |
| IATI Datastore IATI (International Aid Transparency Initiative)Continuous The International Aid Transparency Initiative: an open standard under which donors, UN agencies and NGOs publish their own activity data, often more up to date than the OECD's. Completeness and quality depend on each publisher. | Activities currently in implementation as each funder or implementer publishes them: who is active now, through which implementing organisations, with start and planned end dates | Activity records; no amounts shown or totalled; coverage depends on who publishes | Continuous each publisher on its own schedule (IATI asks for at least quarterly | Live query per brief (API key), cached for six hours per server instance | Funding |
| UN OCHA Financial Tracking Service UN OCHAContinuous The UN's tracking of humanitarian funding against appeals and response plans, from what donors and agencies report. It is the usual source for how well funded an appeal is; recent figures grow as reports come in. | Humanitarian response plans: required vs funded, by year and cluster, with the largest donors | USD millions and % funded | Continuous funding counts once donors or agencies report it | Live query per brief | FundingHumanitarian |
| UN OCHA Country-Based Pooled Funds (CBPF) UN OCHAReal time Funds in individual crises, managed by OCHA, that pool donor money and allocate it to UN agencies and NGOs, including national and local organisations. They show who receives humanitarian money inside the country. | Approved pooled-fund allocations by partner type (UN, international and national NGOs) and the largest funded partners by cluster | USD millions, approved budgets; direct recipients only | Real time from OCHA's grant system, allocations in rounds through the year | Live, cached per server instance | FundingHumanitarian |
| UN Central Emergency Response Fund (CERF) UN OCHAWeekly The UN's global emergency fund, which releases money quickly to UN agencies and IOM for sudden or underfunded crises. Its grants signal where the UN sees urgent or neglected needs. | CERF grants to UN agencies and IOM by country, year and window (Rapid Response, Underfunded Emergencies) | USD millions, approved amounts | Weekly the HDX dataset this app reads | Live, cached per server instance | FundingHumanitarian |
| WFP market prices WFP via the HDX Humanitarian APIMonthly Prices of staple foods that WFP and partners collect in local markets. Rising prices are an early warning of hunger and reduce what cash assistance buys. | Retail prices of staple foods (cereals, pulses, oils) in the markets WFP monitors: latest month and change from a year earlier | Local currency, nominal; median across markets | Monthly the latest month varies by country and market | Live query per brief (food, livelihood and humanitarian scopes); each market's own location and change from a year earlier for the market map | HumanitarianDevelopment needs |
| World Bank World Development Indicators World BankTwo to four updates a year The World Bank's main collection of development statistics on poverty, health, education and the economy, compiled from national and international sources. It is the common reference for country indicators; the latest years are often missing. | Need and context indicators the planner chooses (education, health, energy access, poverty, FDI inflows …) | Indicator units | Two to four updates a year each indicator's latest year differs | Live query per brief | Development needsEconomy and investment |
| Worldwide Governance Indicators World BankYearly The World Bank's composite measures of governance, built from surveys and expert assessments. They are widely used to compare countries, but they measure perceptions and come with margins of error. | Control of corruption, rule of law, regulatory quality, government effectiveness, political stability, voice and accountability | Absolute governance score 0–100, higher is better; not a percentile rank | Yearly the 2025 edition (December 2025) runs to 2024 | All countries fetched once per server instance | GovernanceEconomy and investment |
| IMF World Economic Outlook IMFTwice a year April, October The IMF's twice-yearly economic figures and forecasts for its member countries, the standard reference for growth, inflation and public debt; revenue comes from its Fiscal Monitor. Forecasts are revised as conditions change. | Real GDP growth, inflation, government debt and, from the Fiscal Monitor, government revenue including grants, with projections | % and % of GDP | Twice a year April, October | All countries fetched once per server instance | Economy and investment |
| INFORM Risk Index European Commission JRCYearly An index from the European Commission's Joint Research Centre and partners that rates how likely a humanitarian crisis or disaster is to overwhelm a country's capacity to cope. Humanitarian donors and agencies use it to compare countries; it measures structural risk, not current needs. | Risk that a crisis overwhelms national capacity: hazard and exposure, vulnerability, lack of coping capacity | 0–10, higher is riskier | Yearly plus a mid-year update | Cached per instance; a dated copy of the edition if the API is unreachable | Risk and environmentHumanitarian |
| ACAPS INFORM Severity Index ACAPSMonthly ACAPS's rating of how severe each ongoing humanitarian crisis is, based on its analysts' review of many sources. It is used to compare crises and to see whether one is getting worse. | How severe each ongoing crisis is, its drivers and three-month trend | 0–10 | Monthly | Monthly edition from the Humanitarian Data Exchange, cached | HumanitarianRisk and environment |
| OCHA 3W operational presence UN OCHA via the HDX Humanitarian APIMonthly on HDX 'Who does What Where': OCHA's lists of the organisations that report activities in each area and sector. Useful for finding partners on the ground; it shows only what organisations report. | Which organisations report working in each area, by sector: operational contact leads | Reported geographic areas by administrative level, latest available period | Monthly on HDX each country reports quarterly, yearly or irregularly | Live query per brief (HDX HAPI); organisations per district mapped where the districts match OCHA's boundaries by code and name | Humanitarian |
| People in need (Humanitarian Needs Overviews) UN OCHA HPC Tools via the HDX Humanitarian APIYearly The yearly humanitarian needs overviews agreed by UN agencies and partners in each crisis, with the official number of people in need. They are the basis of humanitarian appeals. | People in need and targeted for the plan year: all sectors (intersectoral) and by cluster where reported | People; sector figures overlap and are never added | Yearly with the humanitarian programme cycle | Live query per brief (HDX HAPI) | HumanitarianDevelopment needs |
| Acute food insecurity (IPC / Cadre Harmonisé) IPC via the HDX Humanitarian APIPer analysis The IPC, and the Cadre Harmonisé in West Africa and the Sahel, are the standard, consensus-based classifications of food insecurity used by governments, the UN and NGOs. Phase 3 means crisis; Phase 5 means catastrophe or famine. | People in Phase 3 (Crisis) or worse, current and projected, for the latest analysis; Cadre Harmonisé in West Africa and the Sahel. Where no analysis covers the present, FEWS NET's latest monthly estimate is added as a range | People and share of the population analysed | Per analysis usually once or twice a year per country, with FEWS NET's monthly estimate where no analysis covers the present | Live query per brief (HDX HAPI), for food, livelihood and humanitarian scopes; district shares mapped only where the analysis uses the same districts as OCHA's boundaries | HumanitarianDevelopment needs |
| ACAPS Humanitarian Access ACAPSAbout every six months ACAPS's rating of how hard it is for humanitarian organisations to reach people in a crisis, because of insecurity, bureaucratic restrictions or physical obstacles. | How hard it is for aid to reach people in each crisis, with its main constraints | 0–5, higher is harder | About every six months monthly API editions repeat the scores in between | ACAPS API, cached | Humanitarian |
| UNHCR refugee statistics UNHCRTwice a year UNHCR's official statistics on refugees and other forcibly displaced people, compiled from governments and its own operations; the global reference for refugee numbers. | Refugees and other displaced people hosted | People | Twice a year end-year figures in June, mid-year figures in the autumn | Live query per brief | HumanitarianDevelopment needs |
| GDACS disaster alerts UN OCHA and European Commission JRCNear real time A UN and European Commission alert system that estimates within hours how serious an earthquake, cyclone, flood or drought may be for people. Alerts are model estimates to guide international response, not damage assessments. | Floods, cyclones, earthquakes and droughts of the last twelve months, with location and alert level | Green, orange, red | Near real time | Live; a circuit breaker says 'unavailable', never 'no alerts' | Risk and environmentHumanitarian |
| ACLED conflict events ACLED (Armed Conflict Location & Event Data Project)Weekly A widely used dataset of reported political violence and protests, coded from news, local partners and other sources. Counts depend on reporting, so areas with little coverage may be under-recorded. | Reported political violence, attacks on civilians and demonstrations, with reported fatalities: latest complete month, twelve months and the twelve before, by country, region and district | Events and reported fatalities; categories overlap | Weekly the app reads monthly totals and leaves out the running month | ACLED's monthly files on HDX, cached for twelve hours; national and regional figures are sums of ACLED's district rows | Risk and environmentHumanitarianGovernance |
| CHIRPS rainfall (WFP) Climate Hazards Center (CHIRPS), aggregated by WFP, via the HDX Humanitarian APIEvery ten days Satellite rainfall estimates combined with weather-station data, produced by the Climate Hazards Center at UC Santa Barbara and widely used in food-security early warning; WFP averages them by region. | Rainfall over the last three months and the last month by region, against the 1989-2018 average, for food, livelihood, water and environment scopes | Millimetres and % of the long-term average | Every ten days recent periods are preliminary and replaced by final data about a month later | HDX Humanitarian API, live query per brief; region names matched to WFP's codes from OCHA's boundaries and checked against each region's size | Risk and environmentDevelopment needs |
| Imputed multilateral aid (ONE, OECD) ONE Data and OECD, via the Development Finance ObservatoryYearly OECD data Estimates of where a donor's core contributions to multilateral organisations end up, by country and sector, by ONE and the OECD. They are modelled from each organisation's spending, not reported project by project. | Norway's core funding to multilateral organisations attributed to each country (OECD DAC2A) and sector (ONE Imputed Multilateral Sectors): modelled estimates | USD millions, gross disbursements, current prices | Yearly OECD data ONE recomputes quarterly | Development Finance Observatory API, fixed queries, cached for a day | Funding |
| IMF-World Bank debt sustainability ratings IMF and World BankWhen reassessed Joint IMF and World Bank assessments of whether low-income countries can carry their debt, rated from low risk to in debt distress. They influence how much of the World Bank's IDA support a country gets as grants rather than loans. | Risk of external debt distress (low, moderate, high, in debt distress) and whether debt is sustainable, for low-income countries, with the date of each assessment | Rating | When reassessed a country's rating changes when a new analysis is published, usually with an IMF programme review, and the IMF updates its list a few times a year | The IMF's list of debt sustainability analyses (PDF), read weekly | Economy and investment |
| World Bank International Debt Statistics World BankYearly each December The World Bank's official statistics on low- and middle-income countries' external debt and debt payments, reported by the borrowing countries. | External debt stocks, their share of GNI and public external debt service | USD millions and % of GNI | Yearly each December | World Bank API, live query per brief | Economy and investment |
| OECD climate-related development finance OECD (recipient perspective: bilateral and multilateral providers)Yearly The OECD's dataset of development finance for climate adaptation and mitigation from bilateral and multilateral providers, the basis of many climate finance studies. Its counting rules make bilateral totals generous. | Climate-related commitments by provider, with adaptation, mitigation and a principal-objective-only total, and the largest climate-related commitments of the latest year, for climate-relevant scopes | USD millions, commitments, current prices | Yearly each year's file arrives about 18 months after the year ends (2024 in June 2026 | The OECD's three newest yearly files, summarised once per release and kept on disk | FundingRisk and environment |
| UNDP human development and IOM displacement UNDP Human Development Report Office (with OPHI for poverty) and IOM Displacement Tracking MatrixYearly UNDP's Human Development Index combines health, education and income into one measure and is the best-known summary of human development; IOM's Displacement Tracking Matrix counts people displaced within their own country in crises. | Human Development Index and rank, inequality-adjusted HDI and the share lost to inequality, Gender Inequality Index, people in multidimensional poverty with the survey year, and internally displaced people where IOM reports them | Index 0–1, %, people; each figure with its own year | Yearly the HDR in spring, the global MPI each October; IOM displacement rounds every few months where IOM works | UNDP's yearly files (newest edition found on its download page) and IOM's series through the UN System Data Commons API, cached for a day | Development needsHumanitarian |
| UNESCO UIS education statistics UNESCO Institute for StatisticsYearly UNESCO's statistics institute is the official keeper of the global education goal's figures: who is out of school, who finishes primary and lower secondary school, and how many children can read and do basic maths by the end of primary. The World Bank's education figures are copied from it, usually later. | Out-of-school children and rates (with girls and boys), completion rates, minimum proficiency in reading and mathematics at the end of primary, and government spending on education, for education scopes | People, % and % of GDP; each figure with its own year | Yearly a main release each September and a refresh of national data each February | UIS public data API, live query per brief; the release name from its versions endpoint | Development needs |
| World Bank projects World BankContinuous The World Bank's own database of the projects it finances, with approved amounts, dates and status, including projects still in preparation. | Active and pipeline World Bank projects with commitment at approval, major sectors and dates | USD millions committed (not disbursed) | Continuous a project appears when it enters the pipeline and is updated at Board approval | World Bank Projects API, live query per brief | Funding |
| Web research AI web searchLive Current news and reports found by AI web search. Shown separately as background and never cited as evidence. | Current reporting on the question, with source links | Text | Live searched at each run | Shown in a separate panel; never cited as evidence |
Country outlines: Natural Earth (public domain). District and region outlines: OCHA common operational boundaries on HDX (CC BY-IGO), drawn only where their codes and names agree with the publisher's, kept on disk for 30 days. Each source is used under its publisher's terms.
● main use · ○ useful context. A brief includes whatever applies to its countries; the reader's purpose decides what leads.
| NGO programmes | Foundations and donors | Humanitarian funds | Social enterprises | Investors and companies | Researchers and officials | |
|---|---|---|---|---|---|---|
| Who funds what, and through whom (Norwegian data, OECD CRS) | ● | ● | ○ | ● | ● | ● |
| Recent aid cuts and budgets (OECD DAC1) | ○ | ● | ● | ○ | ○ | ● |
| Peer foundations and their projects (OECD private philanthropy) | ○ | ● | ○ | ○ | ● | |
| Private finance and guarantees (OECD mobilisation) | ○ | ● | ● | ● | ||
| Who is active now, and through whom (IATI) | ● | ● | ○ | ● | ● | ● |
| What the World Bank finances and plans (World Bank projects) | ● | ● | ○ | ● | ● | ● |
| Norway's money through multilaterals, by country and sector (ONE, OECD) | ○ | ● | ● | ○ | ○ | ● |
| Who funds climate adaptation and mitigation (OECD climate-related development finance) | ● | ● | ○ | ● | ● | ● |
| Humanitarian funding gaps (OCHA FTS) | ● | ● | ● | ● | ||
| Who receives humanitarian money: pooled-fund partners and national NGO share (OCHA CBPF, CERF) | ● | ● | ● | ● | ||
| Staple food prices in local markets (WFP) | ● | ○ | ● | ○ | ○ | ● |
| Rainfall against the long-term average, by region (CHIRPS, WFP) | ● | ○ | ● | ○ | ○ | ● |
| Corruption, rule of law, regulation (WGI) | ○ | ● | ○ | ● | ● | ● |
| Economic outlook (IMF) | ○ | ● | ● | ● | ||
| External debt and debt service (World Bank IDS) | ○ | ● | ● | ● | ||
| Risk of debt distress and government revenue (IMF, World Bank) | ● | ● | ● | ● | ||
| Crisis risk and severity (INFORM, ACAPS Severity) | ● | ● | ● | ○ | ● | ● |
| Access for aid delivery (ACAPS Access) | ● | ○ | ● | ○ | ○ | ● |
| Conflict and protest, down to districts (ACLED) | ● | ● | ● | ○ | ● | ● |
| Who works where: organisations on the ground (OCHA 3W) | ● | ○ | ● | ○ | ● | |
| Scale of need: people in need and acute food insecurity (OCHA HNO, IPC / CH) | ● | ● | ● | ● | ||
| Recent disasters (GDACS) | ● | ○ | ● | ● | ● | ○ |
| Need indicators (World Development Indicators) | ● | ● | ○ | ● | ● | ● |
| Human development, inequality and displacement (UNDP HDR, IOM DTM) | ● | ● | ● | ○ | ○ | ● |
| Out of school, completion and learning (UNESCO UIS) | ● | ● | ○ | ○ | ● |
Plain words in, a validated scope out.
A planner model turns the reader's words into a scope: countries (ISO3 codes), DAC sector codes, indicators, a semantic search query and a web query. The server validates every code against the official lists and drops anything unknown.
31 sources read in parallel; each result becomes a numbered evidence card.
The applicable sources among 31 are read in parallel, live or from a recent cache. From the Norwegian data: exact totals for the scope (totals by year with trend lines, partners, channels, sub-sectors, policy markers) and the largest agreements. From the search by meaning: the agreements whose descriptions are closest in meaning to the reader's purpose (the 14 nearest, 8 kept after removing duplicates of the largest agreements). From the OECD: all donors in the country and sector, and other donors' individual projects matched on purpose codes. Context indices cover up to three countries. Each result becomes a numbered evidence card with its publisher, scope and retrieval date.
The reader's purpose decides what leads, not the volume of data.
Relevance comes from the reader's purpose, not from volume. The planner states the reader's need; the semantic search surfaces agreements that match it in meaning, even with different wording; the composing model is told to lead with what changes the reader's decision, to pair a donor's global trend with its local trend, and to prefer evidence over generalisation.
Written from the evidence cards only; every figure cites one.
One model call writes the brief from the reader's request and the evidence cards; only the cards count as evidence. Every figure must cite its card. It may not make claims about the reader's own organisation, and dramatic wording is replaced by neutral words.
Numbers and names are traced; what cannot be confirmed is flagged, not hidden.
Automated checks look for each cited number in its evidence cards (allowing honest rounding) and flag those they cannot find, confirm citations point to real cards, flag organisations named in no card, flag one-sided trend claims, and remove invented claims about the reader. The checks are heuristics, not proof: they flag some unsupported claims and miss others, so flags are shown to the reader and a person should still review the brief.
Current news stays in its own labelled panel.
A separate search-grounded model reports current news. A comparison step lists where it differs from the brief. It is never mixed into the cited evidence.
Fixed queries find what changed; a model decides whether to write, and says why.
For a watch, fixed queries over the recorded payments (2015–2025, refreshed by an operator) finds the latest-year change in funding, agreements whose first or last recorded payment year changed, and new partners. These are payment records, not contract start or end dates. The same check reads the faster public sources for the scope (OCHA appeal and pooled-fund funding, INFORM Severity, ACAPS access, people in need and IPC, WFP prices, GDACS alerts) and compares their rows with the picture at the last brief; changes that were held add up until they matter. A grounded web search adds reporting since the last brief, as a lead: on its own it can prompt a short update at most. A decision model weighs these signals with the reader's feedback and last brief, and chooses a new brief, a short update or a hold, with a reason. Scheduled checks run weekly; a reader's explicit check or new feedback can bring one forward. Fixed rules (evidence available, the cadence, an email allowlist, the web-only cap) decide whether a brief may be sent at all.
A timer starts the weekly check; the email goes out; feedback shapes the next run.
A timer on the server starts the weekly check of every watch and a daily run for anything due. When the agent decides to send, the app writes the email in the reader's sector colours and sends it through an email service that keeps the list of readers who opted in and handles unsubscribes, bounces and blocks. One-click feedback is stored with the watch and read by the next run. Contact-form messages reach the author by email the same way.
Stated plainly, so results are used for what they can support.
Demonstration prototype: briefs are generated automatically and may contain errors. Quality-assure against the linked sources before use.