What NHCX integration involves
NHCX integration connects a hospital system, insurer or TPA to the National Health Authority’s claims exchange. The hospital sends FHIR bundles for eligibility, pre-authorisation, claims and replies to queries; the payer answers on the hospital’s callback address. Each message is sealed with the recipient’s public key (JWE), addressed with x-hcx- envelope fields and tracked by a correlation ID. It is a separate integration from ABDM milestones M1–M3, with its own participant registration, sandbox exit and production keys.
One integration to reach every payer, if the plumbing is right
Today a hospital’s billing desk keys the same admission into several insurer and TPA portals, and PM-JAY cases into the Transaction Management System on top. NHCX replaces that with one connection: the HMIS builds a claim once, as data, and the exchange routes it to whichever payer holds the policy. The idea is simple. The engineering is not: an encrypted envelope with a dozen headers, a callback address the exchange can reach within thirty seconds, conversations that last days across queries and enhancements, and payer configuration that for PM-JAY runs to tens of megabytes.
We build that layer for HMIS vendors who want NHCX in their product, for hospitals whose vendor has not delivered it, and for insurers and TPAs connecting their claims systems to the payer side. The code lives in your repository, the participant codes and certificates are in your name, and your team can run it after we leave. Radiatus is not an NHA testing agency, a WASA auditor or an NHCX partner of any kind; NHA certifies, independent auditors test, and we prepare the work for both.
NHCX sits beside our ABDM integration work, because PM-JAY on NHCX expects your software to be live on ABDM M1 first. If your product is still on the older ABDM APIs, our ABDM V3 migration guide covers that move.
NHCX and the ABDM milestones are different integrations
Both belong to the Ayushman Bharat Digital Mission and both use FHIR R4 from the NRCeS implementation guide, which is why they get confused. They solve different problems with different plumbing.
| ABDM M1–M3 | NHCX | |
|---|---|---|
| Purpose | Patient identity (ABHA) and consent-based sharing of health records | Insurance: cover, pre-authorisation, claims, queries, payment |
| Counterparty | Other hospitals, labs and PHR apps, through the consent manager | Insurers, TPAs and government schemes such as PM-JAY |
| Identity on the network | HFR facility ID behind your bridge | A participant code such as 1000003538@hcx; for a hospital in production, registered against its HFR ID |
| Message security | Records encrypted between HIP and HIU with ECDH key exchange (Fidelius) | Each bundle sealed as a JWE with the recipient’s RSA certificate from the participant registry |
| FHIR content | Document bundles: discharge summary, diagnostic report, prescription and others | Collection bundles: Claim, ClaimResponse, CoverageEligibilityRequest/Response, InsurancePlan, Task |
| Conversation | Consent ID and transaction ID | x-hcx-correlation_id across every request and answer in a case |
| Go-live | ABDM sandbox exit, WASA, NHA approval | NHCX sandbox exit with NRCeS bundle validation and demos; WASA as well for PM-JAY |
| DHIS incentive (Corrigendum 7) | Per linked record and per consent | Per PM-JAY claim through NHCX: ₹200 or 10% of the claim, whichever is lower, to the hospital; ₹10 to the software company; ₹1 crore cap |
Is NHCX integration mandatory?
Not for most organisations, as of 13 November 2026. What the primary sources say:
- IRDAI, June 2023. Circular IRDAI/HLT/CIR/MISC/124/06/2023 of 8 June 2023 "advised" all insurers and TPAs to test and adopt the NHCX specifications, and said insurers "may on-board themselves" as soon as possible.
- IRDAI, May 2024. The Master Circular on Health Insurance Business (IRDAI/HLT/CIR/PRO/84/5/2024, 29 May 2024) says insurers, with the Insurance Councils, "may endeavor to" onboard hospitals and healthcare providers on NHCX. The same circular’s hard deadlines are about speed: a cashless decision within one hour and discharge authorisation within three hours, which an automated claims channel helps with.
- IRDAI, August 2026. A press release on its health insurance sub-committee meeting of 28 August 2026 (reproduced by TaxGuru) says onboarding strategies and incentives were discussed. No binding rule or date was announced.
- NHA and PM-JAY. NHA recognised its first "NHCX-PMJAY Early Integrators" in March 2026 (PIB, 7 March 2026), reviewed the rollout of the NHCX–HMIS pilot in July 2026 (PIB, 17 July 2026), and the next day described NHCX as a route to faster settlement of approved PM-JAY claims for empanelled hospitals (PIB, 18 July 2026). We found no circular requiring PM-JAY hospitals to move off TMS.
- Money. DHIS Corrigendum 7 pays for PM-JAY claims filed in FHIR through NHCX (April to September 2026), which is an incentive, not an obligation. We found no later corrigendum on 13 November 2026.
So the honest answer: optional today, encouraged by the regulator, paid for by NHA under DHIS while funds last, and clearly where PM-JAY claims are heading. If you sell an HMIS to empanelled hospitals, plan for it; if you are a single private hospital with a handful of insurers, ask your vendor for a date before you fund your own build.
Use cases, API paths and the NRCeS claim profiles
Claim content follows the NHCX profiles in the NRCeS FHIR Implementation Guide for ABDM, version 6.5.0 on FHIR 4.0.1, the current published version on 13 November 2026. Every message is a Bundle of type collection. The guide defines six bundle profiles (ClaimBundle, ClaimResponseBundle, CoverageEligibilityRequestBundle, CoverageEligibilityResponseBundle, InsurancePlanBundle and TaskBundle) and profiles for Claim, ClaimResponse, Communication, CommunicationRequest, Coverage, CoverageEligibilityRequest, CoverageEligibilityResponse, InsurancePlan, PaymentNotice, PaymentReconciliation and Task.
| Step | Provider sends | Answer arrives at | What travels |
|---|---|---|---|
| Get the payer’s plan | /v1/insuranceplan/request | /v1/insuranceplan/on_request | InsurancePlanBundle: packages, tariffs, questionnaires, required documents |
| Check cover | /v1/coverageeligibility/check | /v1/coverageeligibility/on_check | CoverageEligibilityRequestBundle, then the response bundle with benefits |
| Pre-authorise | /v1/preauth/submit | /v1/preauth/on_submit | ClaimBundle with Claim.use = preauthorization; ClaimResponseBundle back |
| Estimate cost | /v1/predetermination/submit | /v1/predetermination/on_submit | ClaimBundle with use = predetermination |
| Claim at discharge | /v1/claim/submit | /v1/claim/on_submit | ClaimBundle with use = claim, bills and supporting information |
| Answer a private insurer’s query | /v1/communication/on_request | The query arrives at /v1/communication/request | Communication with the documents asked for |
| Payment | Acknowledgement on /v1/paymentnotice/on_request | The notice arrives at /v1/paymentnotice/request | PaymentNotice with remittance details |
| Cancel, reprocess, search | /v1/task/submit, /v1/search/submit | /v1/task/on_submit, /v1/search/on_submit | TaskBundle |
| Where is my request? | /v1/status | /v1/on_status | Envelope only |
Paths as listed in the NHCX integration package on GitHub, published under an organisation that names itself the National Health Authority (created August 2026, last updated 25 October 2026). The exchange’s sandbox base is https://apisbx.abdm.gov.in/hcx; check every path against the specification on the NHCX portal before you code.
Workflow codes tell the two sides which step a message is. Examples from the same package: 12 a new pre-authorisation, 20 received and under review, 21 approved, 23 denied, 24 query raised, 19 query answered, 25 to 27 claim acknowledged, approved or queried, and 30 the payment remittance with bank reference and tax deducted.
The envelope: JWE, headers and the conversation ID
Every bundle is sealed for its recipient. The sender looks up the recipient’s certificate in the participant registry, encrypts with RSA-OAEP-256 and A256GCM, and sends the compact JWE string as {"type": "JWEPayload", "payload": "eyJ..."}. The exchange routes on the protected header, which it can read, and never sees the bundle. A protected header looks like this:
{
"alg": "RSA-OAEP-256",
"enc": "A256GCM",
"x-hcx-sender_code": "1000003538@hcx",
"x-hcx-recipient_code": "1000004446@hcx",
"x-hcx-api_call_id": "7d3f0c2e-5b61-4c9a-a1f4-9e2d6b8c0a17",
"x-hcx-request_id": "c0a1e7b4-2f8d-4e3a-9b6c-51d7f2a8e930",
"x-hcx-correlation_id": "7d3f0c2e-5b61-4c9a-a1f4-9e2d6b8c0a17",
"x-hcx-workflow_id": "12",
"x-hcx-timestamp": "2026-11-13T14:20:05+05:30",
"x-hcx-status": "request.initiated",
"x-hcx-ben-abha-id": "91-0000-0000-0000"
}
Four rules cause most early failures:
- Correlation. On a new request the correlation ID equals that call’s
api_call_id; every answer echoes it with a freshapi_call_idof its own. Once a correlation ID fails at the gateway or is stopped after retries, it is dead: retry with new IDs. - Status words.
request.initiatedgoing out,response.partialfor "received, working on it",response.completefor a final answer,response.errorfor a refusal. Sending the wrong one with a workflow code is refused at the gateway. - Timestamp. The package records that NHA’s own documents disagree on format and time zone and that the reference gateways send ISO 8601 in Indian time with
+05:30. Follow the samples and make the format a setting. - Certificates. Encryption fails silently at the other end when the certificate on your participant record is not the pair of the private key you decrypt with. Fetch your own certificate back from the registry after every change and compare.
Callback URL for an on-premise HMIS
NHCX answers arrive at your server, not in the response to your call. The package’s callback chapter sets the rules for the address you register: a domain name over HTTPS with TLS 1.2 or newer, no IP address and no port number, hosted in India, reachable from the exchange’s published outbound addresses, and answering within thirty seconds with HTTP 202 and a receipt. Anything slower counts as a failed delivery; the exchange redelivers up to five times and then reports the failure on your /v1/error endpoint, which you must host too.
Most Indian hospitals run their HMIS on a server in the building, behind a broadband line with no fixed public address. Three designs work:
| Design | How it works | Good for | Watch out for |
|---|---|---|---|
| Cloud relay with an outbound queue | A small service in an Indian cloud region owns the domain, answers the exchange with 202, stores the sealed message and puts it on a queue; the on-premise adapter pulls from the queue over an outbound connection | Hospitals with no inbound firewall rules and unreliable links | The private key must stay on-premise if the relay only stores sealed messages; size the queue for outages |
| Adapter in the hospital DMZ | The adapter runs on-premise behind a reverse proxy with a static public IP, a domain and a certificate | Large hospitals with a network team and a leased line | Uptime of one site; firewall changes; certificate renewals on time |
| Vendor-hosted, multi-tenant | The HMIS vendor runs one callback service for all its hospitals and routes by recipient code | HMIS vendors with many small customers | Tenant isolation; the vendor holds keys for every hospital; contracts must say so |
Whichever you choose, acknowledge first and work second: store the raw body, return the receipt, then decrypt, match the correlation ID and update the case. The package also notes that the exchange signs its calls with an RS256 JWT, but that no public source publishes the key to verify it; ask for that key when you register your endpoint, and meanwhile restrict the endpoint to the exchange’s addresses.
Designing the NHCX adapter layer
We keep NHCX out of the HMIS core, in a service of its own with these parts:
- Participant and key manager: participant code, registry calls, your key pair, cached recipient certificates with expiry.
- Envelope builder: IDs, workflow code, status word, timestamp, ABHA header; one place to change when NHA clarifies a rule.
- Bundle builders and validators: one per use case, mapping HMIS bills, diagnoses, procedures and documents to the NRCeS profiles, validated in CI against the guide’s package.
- Plan store: versioned InsurancePlan responses per payer. For PM-JAY the package’s reference plan is 21 MB with over two thousand questionnaires, so it is data with versions, not a cached blob.
- Case state machine: every pre-authorisation and claim with its correlation IDs, workflow codes, amounts and deadlines, so a query that arrives days later lands on the right case.
- Callback receiver: the thirty-second acknowledgement, idempotency on
api_call_id, raw message archive. - Desk work queue: queries, shortfalls and denials surfaced to the billing team inside the HMIS, with the documents asked for.
- Reconciliation: payment notices matched to claims and bank references, feeding accounts.
NHA’s optional NHCX Adapter, a single Go binary, takes over the token, certificates, envelope, sealing and receipts, and lets the HMIS speak plain JSON and FHIR to it. Its own documentation lists what it does not do: no queueing or retries, no bundle building or validation, and no business logic. It shortens the protocol work; it does not remove items 3 to 8 above.
Provider side and payer side
| Hospital or HMIS vendor | Insurer or TPA | |
|---|---|---|
| Hosts | The on_ callbacks: answers to eligibility, pre-auth, claims; incoming queries and payment notices | The submit and check side: incoming requests from every hospital |
| Main build | Bundles from HMIS data, case tracking, desk work queue, documents | Validation of every incoming bundle, adjudication hand-off, response bundles, plan publishing |
| Volume shape | Tens to thousands of cases a day per hospital | Every network hospital at once; queues and back-pressure matter |
| Credentials | Provider role on the production client after NHA certification | Payer or TPA role after enrolment through the NHA or IRDAI portal |
| Registry ID in production | The hospital’s HFR ID | The IRDAI registration number |
PM-JAY on NHCX: what changes
The package’s PM-JAY chapters, drawn from NHA’s functional requirement document (version 1.0, March 2026, marked work in progress), describe how an empanelled hospital raises PM-JAY cases from its HMIS instead of TMS. The endpoints are the same as for private insurers. Four things differ: the InsurancePlan response carries the scheme’s packages, rates, treatment guidelines and mandatory documents for that hospital; biometric verification of the beneficiary (fingerprint, iris or face) is required before pre-authorisation and again at discharge, through extra APIs; clinical information must be structured rather than scanned; and PM-JAY raises document queries on the pre-authorisation or claim itself, not through the communication API.
Go-live is a manual switch: after production registration, the hospital raises a ticket with its PM-JAY hospital ID, its HEM ID from TMS and its new NHCX participant ID, and NHA maps them. Cases opened before the mapping finish in TMS, so for a while the desk runs both.
From sandbox to production
- ABDM first. For PM-JAY, the integrator journey starts with going live on ABDM M1, with its own functional testing and WASA audit.
- Register on the NHCX sandbox with your ABDM client ID, create a participant code, generate a key pair and register your certificate and callback address.
- Prove the loop against the dummy payer: send a sealed bundle, trigger an approve, reject or query, and receive it on your callback.
- Build every use case on your side’s checklist, including cancel, reprocess, status and duplicate callbacks, not just the happy path.
- Exit the sandbox: sample bundles emailed to NHA for NRCeS validation, an internal demo to NHA, a demo to the Health Tech Committee and, for PM-JAY, a PM-JAY team demo, a WASA audit and the exit form.
- Production: NHA adds the provider role to your production client; you register the participant again in production through a passcode flow sent to the mobile number on your HFR record, then register the production certificate and callback address.
- Pilot a few real cases, train the desk, then switch fully; for PM-JAY, raise the mapping ticket.
Radiatus does not certify or audit anything in this list. We prepare the code, the bundles and the evidence, and sit with your team through the demos.
NHCX integration services
Scoped after a short specification review and priced per phase. Participant codes, certificates and production credentials stay in your organisation’s name.
Provider-side NHCX for HMIS vendors
The problem
Hospitals want cashless claims from your product, and the protocol work keeps slipping behind the roadmap.
How we do it
We build the adapter layer as a separate service in your repository: envelope and sealing, registry and certificates, bundle builders and validation, case tracking, callbacks and the desk work queue, with tests mapped to the provider checklist.
What you get
- NHCX service with tests
- Bundle builders for each use case
- Desk work queue in the HMIS
- Exit evidence and runbook
PM-JAY on NHCX
The problem
Your hospital or your HMIS customers raise PM-JAY cases in TMS and want them in the HMIS.
How we do it
The PM-JAY extras on top of the provider module: the scheme’s plan configuration, biometric authentication flows, structured supporting information, in-case queries, and a planned cutover with the mapping ticket.
What you get
- Plan store and package selection
- Biometric authentication flows
- Structured supporting data
- Cutover plan for TMS to HMIS
Callback hosting for on-premise HMIS
The problem
Your HMIS sits in the hospital and the exchange cannot reach it, or reaches it too slowly.
How we do it
A small relay in an Indian cloud region that owns the callback domain, acknowledges within the time limit, stores sealed messages and hands them to the on-premise adapter over an outbound connection, with monitoring and alerts.
What you get
- Relay service and queue
- Domain, TLS and allow-listing
- Outbound connector on-premise
- Monitoring and alerting
Payer-side NHCX for insurers and TPAs
The problem
You need to receive from every network hospital and answer in the protocol’s shape without rewriting your claims system.
How we do it
A payer-side adapter that validates incoming bundles, returns protocol errors promptly, hands clean cases to your existing adjudication system and builds response bundles and payment notices from its decisions.
What you get
- Payer-side adapter
- Validation and error responses
- Adjudication hand-off
- Payer checklist dry run
Where NHCX projects usually start
Illustrative starting points, not client case studies.
| Organisation | Starting point | Scope we would usually propose |
|---|---|---|
| HMIS vendor selling to PM-JAY hospitals | Customers ask for PM-JAY from the HMIS; product is live on ABDM M1 | Provider-side NHCX module with PM-JAY plan store, biometrics, structured supporting data and the mapping cutover |
| Multi-specialty private hospital | Billing desk re-keys into several insurer and TPA portals; HMIS on a local server | Cloud relay for callbacks, adapter beside the HMIS, eligibility, pre-auth and claim for the insurers already live on NHCX |
| TPA | Payer-side endpoints built, but hospitals report silent failures | Callback and certificate audit, correlation handling, validation responses and monitoring |
| Insurer starting on NHCX | Claims system with no exchange connection | Payer-side adapter, bundle validation, plan publishing and hand-off to the existing adjudication system |
Typical effort for a provider-side NHCX integration
Illustrative engineer-day ranges for a web HMIS with an API layer and an existing ABDM M1 integration, confirmed after a specification review. They exclude NHA’s review time and any auditor’s work.
Foundations
Participant registration, keys, session and registry calls, envelope builder, sealing and opening.
Callbacks
Receiver with receipts and idempotency, plus a cloud relay if the HMIS is on-premise.
Core use cases
Plan, eligibility, pre-authorisation with enhancement and queries, claim, payment notice, status and tasks.
PM-JAY extras
Plan store, biometric flows, structured supporting information, in-case queries.
Desk and reconciliation
Work queue inside the HMIS and payment matching.
Checklist and exit
Dry runs of the checklist, bundle samples for NRCeS, demos, WASA fixes.
Specifications and sources we build against
Read on 13 November 2026. NHCX documents are still being revised, so we recheck them at the start of every project.
NRCeS FHIR IG for ABDM 6.5.0, NHCX profiles
Bundle and resource profiles for eligibility, plans, claims, communications, tasks and payments.
NHCX integration package (GitHub, nha-in)
Base URLs, envelope rules, callbacks, checklists, PM-JAY chapters and sample bundles; published under an organisation that names itself NHA, created August 2026.
NHCX Adapter (GitHub, nha-in)
Optional Go binary that handles tokens, certificates, sealing and receipts.
IRDAI circulars of 2023 and 2024
The advice to adopt NHCX and the cashless timelines in the Master Circular.
DHIS Corrigendum 7
Rates for PM-JAY claims through NHCX, April to September 2026.
PIB releases, March and July 2026
Early integrators, the NHCX–HMIS pilot and settlement of PM-JAY claims.
Tools we use alongside it
- HAPI FHIR
- HL7 FHIR Validator
- Nimbus JOSE + JWT
- jose (Node.js)
- Bruno
- PostgreSQL
- RabbitMQ
- Docker
Facts, versions and vendor details checked 13 November 2026.
Guides and related pages
Guides
Questions about NHCX integration
The National Health Claims Exchange is the claims gateway of the Ayushman Bharat Digital Mission, run by the National Health Authority. Hospitals, insurers, TPAs and government schemes connect to it once and exchange eligibility checks, pre-authorisations, claims, queries and payment notices as encrypted FHIR bundles, instead of each hospital integrating with each payer.
Not as of 13 November 2026. IRDAI advised insurers and TPAs in June 2023 to test and adopt NHCX, and its May 2024 Master Circular says insurers may endeavor to onboard hospitals on it. An IRDAI committee discussed onboarding incentives in August 2026 without announcing binding rules. NHA is piloting PM-JAY claims through NHCX and pays incentives for them under DHIS.
The ABDM milestones handle patient identity and consent-based sharing of health records between providers. NHCX handles insurance transactions between providers and payers. They use different FHIR profiles, different encryption, different registration and separate sandbox exits, although PM-JAY on NHCX expects your software to be live on ABDM M1 first.
The NHCX profiles in the NRCeS FHIR Implementation Guide for ABDM, version 6.5.0 on FHIR R4 as of November 2026: bundles for claims, claim responses, coverage eligibility requests and responses, insurance plans and tasks, plus profiles for Claim, ClaimResponse, Coverage, Communication, CommunicationRequest, InsurancePlan, payment notices, PaymentReconciliation and Task.
Yes, but not directly from most hospital networks. The callback must be a domain name over HTTPS with TLS 1.2 or newer, hosted in India, without an IP address or port, and must answer within 30 seconds. Most on-premise hospitals use a small relay in an Indian cloud region that acknowledges the exchange and passes sealed messages to the HMIS over an outbound connection.
Yes, for hospitals whose HMIS has been integrated and mapped. NHA recognised its first NHCX-PMJAY early integrators in March 2026 and reviewed the NHCX-HMIS pilot in July 2026. PM-JAY adds biometric verification of the beneficiary, scheme plan configuration, structured clinical data and a manual mapping step at go-live, after which new cases move from TMS to the HMIS.
DHIS Corrigendum 7, covering April to September 2026, pays the hospital Rs 200 or 10 percent of the claim amount, whichever is lower, for each ABHA-linked PM-JAY claim filed in FHIR through NHCX, and Rs 10 per claim to the software company, capped at Rs 1 crore each. We found no later corrigendum on 13 November 2026.
For a web HMIS with an API layer and ABDM M1 already live, a provider-side integration of the core use cases typically takes about 45 to 90 engineer-days, and PM-JAY adds roughly 15 to 30 more. Calendar time depends on NHCX sandbox access, NRCeS bundle validation, the demos and, for PM-JAY, the WASA audit.
Planning NHCX for your HMIS, hospital or claims system?
Tell us which side you are on, how your system is hosted, which payers or schemes matter, and where you are in the NHCX sandbox. A senior engineer will reply with the scope we would propose and a typical effort range, confirmed after a specification review. Or book a free 30-minute call.






Compliance-first delivery