How this site counts.

It is the number everyone wants, and it is the number that cannot honestly be produced. Indian assistance takes at least seven distinct forms, held in different records, measured in different units, at different stages of completeness. Adding them produces a headline that no official document supports.

So this site reports them separately, and never adds them:

• Grant assistance — money given, not repaid.
• Loans and Lines of Credit — money repaid, with interest.
• Humanitarian assistance — measured in tonnes, doses, flights, personnel.
• In-kind supplies — goods, counted as goods.
• Development projects — counted as projects, with their funding type attached.
• Capacity building — training places, counted as places.
• Multilateral contributions — money routed through UN agencies and funds.

When you see a total on this site, it is labelled with exactly what it sums: which instrument, which stage, which years.

Indian government spending passes through several accounting stages, and the same rupee appears at each of them. A figure is meaningless without knowing which stage it comes from. Every financial record here carries exactly one stage, and records at different stages are never summed.

The practical consequence: this site will not tell you India “spent” an amount that appears only as a Budget Estimate. If the actual expenditure has not been published yet, the page says so rather than borrowing the estimate and calling it spending.

The single most common distortion in reporting on Indian assistance — in both directions, by supporters and critics alike — is treating credit as aid. Every financial record here carries an instrument, and repayable instruments are labelled REPAYABLE FINANCING wherever they appear.

When India flies relief material into a disaster zone, the government usually announces what was sent — tonnes of supplies, numbers of flights, size of the rescue team — and usually does not announce what it cost. Converting that consignment into rupees would require us to price the goods, price the airlift, and pick an exchange rate. Every one of those choices would be ours, not the government's, and the resulting figure would be an invention wearing the clothes of a fact.

So this site does not do it. Where an official monetary value exists, it is shown and attributed. Where it does not, the page says “monetary value not publicly disclosed” and shows the physical quantity — which is, in any case, the more vivid number.

The humanitarian data model is built around this. An assistance item carries a quantity and a unit from an open vocabulary — tonnes, doses, flights, sorties, ships, personnel, field hospitals, bridges, generators, houses, megawatts — and a monetary value only as an optional field that stays empty unless a source filled it.

India runs two kinds of operation that arrive in the same language. In one, India gives something to another country: relief material, medicine, a field hospital, engineers. In the other, India evacuates its own nationals from a place that has become dangerous. Both are humanitarian, both are official, both are announced the same way — and only the first is foreign assistance.

The difference is not pedantry. Operation Kaveri's published expenditure of ₹45.63 crore was the cost of bringing 4,097 Indians out of Sudan. Counted as assistance to Sudan it would be the largest single overstatement available to this site, and it would look entirely plausible.

So every operation is classified by what it actually contains, never by its name:

• International assistance — nothing consular recorded.
• Mixed — both. Its assistance counts; its evacuation does not.
• Evacuation / consular — everything recorded is consular. Kept in full on the site, and counted towards nothing.

The rule works per item, not per operation, which is what makes a mixed operation honest: Operation Maitri's relief to Nepal counts, and the civilians it flew out of Nepal do not. Transport takes its meaning from the operation around it — ships and aircraft carrying relief in are assistance; the same ships and aircraft carrying people out are the means of an evacuation.

Foreign nationals evacuated alongside Indians are shown, because India did carry them. That does not convert the evacuation into aid to the country they left.

Nothing is deleted or hidden by this. Reclassification is recorded in the revision trail, the operations remain in the record with their sources, and the downloadable data carries the distinction as operation_category and counts_as_assistance so anyone re-aggregating it reaches the same answer.

While a relief operation is running, the government restates a running total. Read carelessly, three announcements look like three deliveries:

Adding those figures gives 127 tonnes. The true answer is 68.5 tonnes — the running total the government published, which already contains the consignments above it. Overstating India's help by 85 per cent would not flatter India; it would hand any critic an easy way to discredit the whole site.

So every assistance figure is stored with what kind of figure it is — cumulative, incremental or unknown — and the date it was true as of. The headline is always the latest cumulative claim. Earlier claims are kept and shown as a progression, never summed. The same rule applies to tonnes, doses, flights, personnel, money and anything else that gets restated.

A Line of Credit is extended through the Export-Import Bank of India with Government of India support. The partner government borrows, spends the money largely on Indian goods and services, and repays with interest on concessional terms. It is a real instrument of Indian influence and development. It is not aid.

Each credit line here tracks approved, operative, disbursed, outstanding and annulled amounts separately, because:

• An approved amount may never become operative.
• An operative amount may be only partly disbursed.
• An annulled amount was never spent and must not appear in a total.

Two categories are flagged and held apart. Interest-capitalisation credit lines, extended so that unpaid interest can be rolled into new debt, are refinancing rather than new development assistance and are excluded from development totals. Defence-related credit lines remain in the database, because omitting them would be its own distortion, but they are classified separately and never presented as development assistance.

Government expenditure is recorded in Indian financial years running 1 April to 31 March. Events happen on ordinary calendar dates. Conflating the two is a common source of error, so every record stores both: the real event date, and the financial year the spending falls in.

An earthquake response beginning on 29 March therefore sits in one financial year while a response beginning on 2 April sits in the next, even though they are four days apart. The site shows the event date by default and the financial year where the accounting matters.

The tracked period runs from 1 April 2000 to the present day.

This site updates itself. An automated pipeline checks official sources on a schedule and decides what happens to every claim it finds, using deterministic rules — not a confidence score, and not a person clicking Approve. There are three outcomes:

Money is held to a higher bar than physical quantities. A rupee figure publishes automatically only when it comes out of a schema-stable official table — the Union Budget Demands for Grants, the Exim Bank register — and reconciles against a printed total in the same document. A monetary claim inside a press-release sentence is prose, and prose is where misreadings live, so it is held.

Sources are ranked by publisher, not by how confident anything felt. Tier A is the Union Budget, CAG, Parliament, MEA and the Exim Bank. Tier B is PIB, Indian missions, the NDRF and other ministries. Media is Tier C: it can point the crawler at an official document, but it can never be the evidence for a figure.

Nothing is ever labelled “verified” because it is pending. A held claim is shown as held. A repository-only citation is shown as a repository, not as an exact source.

If a figure here is wrong, we want to fix it, and we would rather hear it from you than defend it. Every record keeps a revision history, and a corrected record shows what changed and when.

The strongest correction is a link to a primary Government of India document that says something different from what this site says.