Interoperability With Existing eGov Stacks

Every second state AI RFP now contains a sentence like 'integrate with DigiLocker, UMANG, Aadhaar, the state service bus and all department APIs'. That sentence is not a specification. It is a wish list that forces bidders to invent endpoints. The winning SI then spends a year discovering that DigiLocker is a consent-bearing document rail, UMANG is a catalogue, Aadhaar is a UIDAI product, and the state bus is a tired ESB whose owner has left.

Interoperability is a guest discipline. You ask for a published connector. You name a purpose. You log the call. You do not scrape, you do not hoard, and you do not describe a fantasy API as if MeitY had already built it for you.

This guide is for programme SIs and state architects. It is dated 17 August 2026. It is not legal advice and it is not a substitute for the live DigiLocker, UMANG or state-bus manuals. Those manuals change. Read them.

Name the object you are actually calling

DigiLocker holds issued documents a resident or an issuer can share through defined flows. It is not a government Dropbox you may crawl. If you need an income certificate, the clean path is usually: the resident presents it, or your department is an issuer / requester under the current rules. An agent that 'pulls all lockers for the village' is not integration. It is a proposal for a leak.

UMANG aggregates citizen services. Your job is to be a well-behaved service, or to deep-link to one, not to colonise the app with an unconstrained agent. If you want a conversational front end, host it where GIGW, identity and logs are yours, and let UMANG point at it if the programme allows.

State service buses vary. Some are mature. Some are a folder of SOAP endpoints with a 2014 WSDL and a phone number. Write what exists. A purpose-tagged connector to one working endpoint beats a diagram with twelve boxes labelled 'API'.

How to describe eGov touches without inventing a platform.

Stack	Honest description	Invented description to delete
DigiLocker	Connector via published issuer / requester / partner path; purpose tag per fetch	Bulk pull of citizen lockers for training
UMANG	Service listing or deep link if accepted	Embedded general-purpose agent inside UMANG
Aadhaar	UIDAI authentication or e-KYC as permitted; vaulted references	Aadhaar number as the department primary key
State service bus	Named endpoint, owner, SLA, purpose tag	All departments, real time, undocumented
Department portal	Published API or retrieved PDF circular	Screen-scrape behind a clerk login

Purpose tags are the integration

A connector without a purpose tag is how personal data starts to roam. The tag should be short enough to log and strict enough to audit: scheme, file number, document type, whether the resident initiated the share. DPDP purpose limitation becomes visible here. If you cannot write the tag, you cannot make the call.

Keep copies mean. If DigiLocker already holds the issued PDF, prefer a fetch-at-need over a departmental lake of certificates. Lakes become stale, leak, and get embedded. Need is a better architecture than hoard.

Two integrations after the same architecture slide

Objections you will hear — and what to do with them

The RFP already promised all integrations

Corrigendum it. A dishonest RFP is how you buy a dishonest SI. This cluster's API-gaps teardown is the longer version of that sentence.

We will use the sandbox key in production for a while

No. Sandbox keys and production personal data are how you become a case study for the wrong reasons.

Citizens expect one chat that does everything

They expect a service that works. A chat that pretends to fetch a locker document it cannot fetch is cruelty. Refuse, and point to the real locker share.

Purpose tags will slow the bus

A string in a log is cheaper than a purpose-limitation finding. If the bus cannot carry a header, log the tag on your side and still send only what the purpose needs.

A four-week playbook for an honest interface schedule

1. Week 1: inventory claimed integrations. For each, attach a URL or a letter. If neither exists, move the row to 'not in this contract'.
2. Week 2: write purpose tags. Kill any call that needs a tag you are ashamed to log.
3. Week 3: implement one real connector in a test tenancy. Packet-capture it.
4. Week 4: publish the interface schedule as an annex the SI will be paid against. Scrapes are not annex rows.

File note you can paste

Subject: Interoperability of the agent with existing e-governance stacks — connectors and purpose tags only.

Integrations will be listed as connectors to published or lettered interfaces. Each call will carry a purpose tag and will be logged. This department will not invent APIs for DigiLocker, UMANG, Aadhaar or the state service bus. Screen-scraping and shared portal passwords are not integrations. Copies of locker documents will be fetch-at-need unless a written reason requires a store. Citizen-facing surfaces will consider GIGW 3.0.

This note is an internal aid. It is not legal advice.

Prcept AI will integrate with what exists. We will not bid a fantasy bus. If that loses a marking scheme that awards points for imaginary endpoints, the marking scheme should lose, not the file.

When the only interface is a PDF

That is still an interface. Retrieve the PDF, hash it, date it, and draft from it. An SI who treats a PDF circular as failure is an SI who needs a story. Many line departments will publish PDFs for years after they promise APIs. Your clerk can live on that diet if you let it.

Version the PDF store. Last week's labour checklist and this week's labour checklist are not the same object. Completeness against the wrong week is how investors are asked for abolished annexures. The agent should refuse if it cannot see a dated, in-force artefact. Refusal is interoperability too.

If a department later offers a real endpoint, add a connector row. Do not keep the scrape 'as backup'. Backups that still have passwords become the production path at 2am. Kill the guest when the host arrives.

This article is informational field guidance for Indian public institutions, not legal, platform or procurement advice. Confirm against live DigiLocker, UMANG, UIDAI and state-bus documentation, DPDP, GIGW 3.0, and counsel before you file it.

How to sequence this in a state, not a slide

“Interoperability With Existing eGov Stacks” is a department problem. A P4 System Integrator should name the legacy system, the officer who owns the file, and the citizen charter clock before buying “eGov interoperability AI”.

DigiLocker, UMANG and state service buses are real. Your agent is a guest. Describe each touch as a connector and a purpose tag. If the platform has not published an API, you do not have one. Do not invent league tables of states. Read tenders and policies. Election Model Code of Conduct can freeze a rollout. NIC is a partner, not a villain. SDC readiness is GPU, power, ops and egress — not a logo.

• Audit the legacy store first.
• Keep mutation and money as officer actions.
• Map SLAs to the citizen charter.
• Budget change requests after go-live.

Close this loop before the next CAB

Put “Interoperability With Existing eGov Stacks” on the next change-advisory or bid-opening agenda as a single line item with an owner. If it cannot earn a line item, it will not earn a control. The owner should be a P4 System Integrator, not “the vendor.”

Revisit the item when the model, the GeM term, the region, or the SI changes. “eGov interoperability AI” is not a one-time workshop. It is a watch item. Date the last check. Unsigned watch items are souvenirs.

What must be true before you file this

If “Interoperability With Existing eGov Stacks” is only a heading, it will not survive a file inspection. A P4 System Integrator should be able to attach one artefact that proves “eGov interoperability AI”: a log export, a clause, a scored row, a dated notice, or a refusal rule.

Write three dated sentences: what was decided, who owns it, and when it will be re-checked. Unsigned sentences are souvenirs. Dated sentences are controls.

• Name the owner of “eGov interoperability AI” inside the institution.
• Attach one artefact a stranger can open next year.
• Revisit when the model, the notice, or the SI changes.
• Do not treat a vendor slide as evidence.