Key Takeaways
What Epic integration actually requires, which FHIR version to build against, and why the health system rather than the vendor is the gate.
- Epic’s own developer site is free to join. Joining it does not get you into any hospital’s environment, which is a separate approval entirely.
- HL7 publishes FHIR R5 as current, but US certified health IT is required to support FHIR Release 4.0.1. Build against the regulated version.
- Certified EHRs must expose a standardized read API under 45 CFR 170.315(g)(10). Writing data back is not part of that requirement.
- The information blocking rule names its actors in 45 CFR 171.103, and health IT developers of certified health IT are one of them.
- Treat the integration as the last milestone in the pilot, not the first. Everything before it can be built and demonstrated without it.
You are integrating with one hospital’s instance, not with Epic
This is the distinction that resets the whole project plan, and almost every founder gets it the wrong way round on first contact.
Quick Verdict: Register on Epic’s developer site today, because it is free and it tells you which resources exist. Then treat the real work as the health system’s approval process, not the API. Build the product so it demonstrates value without live data first, and add the integration as the final milestone. If you are handling protected health information at all, the architecture decisions in a HIPAA-compliant application build come before the integration decisions, and a regulated web application build should sequence them in that order.
Epic publishes a developer program called Epic on FHIR. Its own site describes it as a free resource for developers who create apps for use by patients and healthcare organizations, and says of any health system, hospital, or clinic running its health record system that it “can connect to any app”.
Read that clause closely. It says a health system can connect. It does not say the health system has done so, nor what connecting involves on their side. The site offers client registration, described as software registration and client identifier management, and a testing sandbox for testing APIs against example data. None of that is access to a live patient record.
So the mental model to carry into your pilot conversation is this: Epic supplies the standard and the plumbing, your pilot hospital supplies the decision. Their IT organization, their security team, their compliance office and their interface analysts are the path.
That also means each additional health system is a new project. Different build, different enabled resources, different approval queue, different go-live calendar. Multi-customer EHR work resembles multi-tenant architecture more than it resembles adding a second API key.
Do this: Ask your pilot contact who owns third-party app approval at their organization, and get that person’s name this week.
Build against the regulated FHIR version, not the newest one
Here is the trap that costs real weeks, and it is created by two sources both being correct.
HL7’s FHIR site currently labels itself FHIR v5.0.0, Release 5, and says plainly that this is the current published version. The label on it is R5 as a Standard for Trial-Use.
Now look at what US regulation requires. The certification criterion at 45 CFR 170.315(g)(10), the standardized API for patient and population services, points at 45 CFR 170.215(a)(1), which names HL7 FHIR Release 4.0.1.
| Source | Version named |
| HL7’s own site, current published version | FHIR v5.0.0, Release 5, Trial-Use |
| US certification criterion 170.315(g)(10) | FHIR Release 4.0.1 |
| US Core Implementation Guide required | STU 6.1.0, with STU 7.0.0 and STU 8.0.1 available through the advancement process |
| Epic on FHIR resource list | R4, STU3 and DSTU2 |
Epic’s resource summary marks its resources R4, STU3 or DSTU2. R5 does not appear.
So a founder who reads that R5 is current, and builds against R5, has built against a version neither the regulation nor the vendor is working in. My read: in US healthcare, treat the certification criterion as the version authority and HL7’s release page as background.
The target also moves on purpose. US Core is required at STU 6.1.0, with STU 7.0.0 from May 2024 and STU 8.0.1 named as approved alternatives, which is the Standards Version Advancement Process at work. The data set those guides carry, the United States Core Data for Interoperability, has published versions through V6 in July 2025 with V7 in draft.
Do this: Write the exact version string you are building against into your technical spec: FHIR 4.0.1 and a named US Core STU version. Confirm it with the health system’s analyst before you write code.
Know what the certified API must give you, and what it need not
The regulation guarantees you a read. It does not guarantee you a write. That single asymmetry decides most product roadmaps in this space.
The United States Core Data for Interoperability is, in the publisher’s words, a standardized set of health data classes and constituent data elements for nationwide, interoperable health information exchange. The g(10) criterion is what makes those elements reachable through a standardized API on certified systems.
| What you want | Where it stands |
| Read standardized patient data | Covered by the g(10) standardized API on certified health IT |
| Read a data element outside USCDI | Not guaranteed. Depends on the instance and what is enabled |
| Write a note, order or flowsheet row back | Not part of the g(10) read requirement. A separate negotiation |
| Launch inside the clinician’s workflow | Possible through SMART App Launch, subject to the health system enabling it |
| Bulk export for a population | Different mechanism and different approval than single-patient reads |
Teams underestimate row three by an order of magnitude. Writing into a clinical record touches patient safety review, not just an API scope, and a health system will treat it that way.
Practically, this means your first version should prove its value on reads. If your product only works by writing back, you have a longer pilot than your runway probably assumes, and you should know that in week one rather than month five. The integration layer itself is ordinary work once scoped, closer to any third-party API integration than to anything exotic.
Do this: Split your data requirements into a read list and a write list, then check whether version one can ship with the read list alone.
Information blocking is the lever, and it names its actors
This rule is why standardized reads are available at all, and it is worth understanding because it cuts both ways.
The Office of the National Coordinator for Health Information Technology defines information blocking as a practice by an actor that is likely to interfere with the access, exchange, or use of electronic health information, except as required by law or specified in an information blocking exception. The exact regulatory definition sits in 45 CFR 171.103, and the exceptions are identified across 45 CFR Part 171.
The actors matter to you. The Cures Act applied the law to healthcare providers, health IT developers of certified health IT, and health information exchanges or health information networks.
Two consequences for a founder. First, a flat refusal to share data that a patient or provider is entitled to is not simply a commercial decision for the organizations named as actors, which is part of why the standardized read exists. Second, if your own product ever becomes certified health IT, you are on that list too.
What the rule does not do is oblige a health system to adopt your application, give you a production client ID, or connect you to anything. Access to data is not the same as access to a customer, and conflating the two is the most common misreading of the Cures Act I see in pitch decks.
Do this: Read 45 CFR 171.103 once, then stop. Have counsel brief you on the exceptions rather than reasoning from the summary yourself.
Implement SMART App Launch exactly as the criterion names it
If your app launches from inside the EHR, the authorization details are specified rather than left to you, and the specification is unusually precise.
The criterion points at 45 CFR 170.215(c)(2), which names the HL7 SMART App Launch Implementation Guide Release 2.0.0. It requires mandatory support for all capabilities defined in the guide’s section 8.1.2, with the permission-online capability excepted, and it requires token introspection as defined in the guide’s section 7.
Alongside it, 45 CFR 170.215(e)(1) names OpenID Connect Core 1.0 incorporating errata set 1, for the authorization paths that need identity.
The test procedure expects three launch modes to work: EHR-Launch, Standalone-Launch, and both. It also expects a discovery document at the .well-known/smart-configuration path.
GET https://<fhir-base>/.well-known/smart-configuration
Accept: application/json
That single request is the cheapest diagnostic available to you. It tells you which capabilities a given instance advertises before anyone schedules a meeting.
One dated detail worth holding: the criterion’s patient authorization revocation requirements, at paragraph (g)(10)(vi), are stated as required by 11 March 2024. Those are past, so expect them present rather than pending.
Do this: Request the SMART configuration document for your pilot’s FHIR endpoint and compare what it advertises against the capability list you need.
The signatures you need are not Epic’s
Access is a paperwork problem wearing a technical costume, and the paperwork belongs to the health system.
Expect at least four things, in roughly this order, and expect each one to have its own owner and its own queue.
| What | Who signs or approves | What it gates |
| A business associate agreement | The health system’s legal or compliance office | Whether you may handle their patient data at all |
| A security review or questionnaire | Their information security team | Whether your architecture is acceptable |
| Internal sponsorship | A clinical or operational owner inside the organization | Whether anyone there is accountable for the project |
| Instance configuration and a production client ID | Their Epic analysts and IT leadership | Whether your app can reach their environment |
The security review is the one that catches unprepared teams, because it asks for evidence rather than intentions. Encryption at rest and in transit, access control, logging, incident response, vendor subprocessors, data retention. If you have already done the work behind a SOC 2 program, most of that questionnaire is already answered, which is the real reason it is worth starting early rather than after a hospital asks.
Internal sponsorship is the one nobody tells you about and the one that actually decides timelines. An enthusiastic clinician who wants your product will move your request through four departments. Without one, your ticket sits.
Do this: Ask your pilot contact directly whether they are willing to be named as the internal sponsor. Their answer tells you whether this pilot is real.
Sequence the pilot so the integration lands last
Almost everything you need to prove can be proven before a single live record moves.
Build the product against Epic’s testing sandbox and synthetic data. Demonstrate the full workflow, the clinician-facing screens, the outputs and the reporting. Run the security questionnaire in parallel with development rather than after it. Start the business associate agreement the week the pilot is verbally agreed, because legal review is calendar time you cannot compress.
Then connect. By the time you do, the only open question is whether the data flows, which is the smallest risk in the whole plan.
Teams that invert this spend the first two months waiting for access they could have waited for while building, and then discover a product problem with no runway left to fix it. Scoping a pilot this way is the same discipline as scoping any first version, and the cost drivers of a first build sit in the same place: how much you commit before you have learned anything.
Do this: Draw the pilot as a timeline with the integration as the final milestone, then start the BAA and the security review on day one of that timeline.
What this primer cannot tell you
Some of this is specific to your organization and your states, and no article resolves it.
I cannot tell you which FHIR resources your pilot’s instance has enabled. That is a per-instance configuration question and only their analysts can answer it.
I cannot tell you their approval timeline. It varies by organization and it is not published anywhere.
I also cannot tell you whether your particular use case needs write access, bulk export, a different authorization path, or a data use agreement on top of the business associate agreement. Nothing here is legal, compliance or regulatory advice. Work with healthcare counsel and with a privacy officer, and treat your pilot’s own security and compliance teams as the authority on what they will accept.
One honest limit on this piece itself. Epic’s public developer site says that any organization running its record system “can connect to any app”, but it does not document what that connection requires on the health system’s side, and I could not find that documented publicly. The approval steps described above are the pattern reported consistently across health-tech teams, not a published Epic process, and your pilot’s actual process may differ. Ask them rather than trusting this list.
Where we fit: if you have a pilot hospital and need the product built so the integration is the last milestone rather than the first blocker, the first step is Get your build plan. It is complimentary, comes back in two business days, and includes the recommended scope, the stack with reasons, the cut list, a fixed price and a launch date. No sales call. We are the wrong call if you need signed HIPAA and SOC 2 attestations at launch, because that is a longer program than a first build and we will say so rather than sell it. Get your build plan and we will tell you which of those two you are. You can see how we scope and sequence work and read what previous builds produced first.
Do this: Confirm your pilot’s FHIR version and enabled resources in writing before committing to a launch date.
Frequently Asked Questions
How much does Epic integration cost?
Epic’s developer program, Epic on FHIR, describes itself as a free resource for developers. Your costs sit elsewhere: engineering time against FHIR 4.0.1 and a named US Core version, the security questionnaire and any remediation it triggers, legal review of the business associate agreement, and the calendar time of the health system’s approval queue.
Which FHIR version should I build against?
FHIR Release 4.0.1, unless your pilot tells you otherwise in writing. HL7 publishes R5 as the current version, but US certification criterion 45 CFR 170.315(g)(10) names Release 4.0.1, and Epic’s published resource list shows R4, STU3 and DSTU2 rather than R5. Confirm the US Core STU version with the health system’s analyst.
Can I write data back into Epic?
Not through the standardized read API that certification requires. The g(10) criterion covers standardized access to USCDI data, and writing into a clinical record is a separate conversation with each health system that typically involves patient safety review rather than only an API scope. Plan version one around reads.
Does the information blocking rule mean a hospital has to work with me?
No. The rule, defined at 45 CFR 171.103, addresses practices that interfere with access, exchange or use of electronic health information by named actors. It does not oblige any organization to adopt your application, issue you a production client identifier, or connect you to their environment. Data access and customer access are different things.
How long does getting Epic access take?
There is no published timeline, and it varies by health system rather than by Epic. The variable that moves it most is whether you have an internal sponsor who wants the project. Start the business associate agreement and the security review on day one of the pilot and build against the sandbox while they run in parallel.
Resources
- Epic on FHIR, Epic Systems. The developer program description, client registration, the testing sandbox, the resource summary marked R4, STU3 and DSTU2, and the statement that an organization running its record system can connect to any app.
- § 170.315(g)(10) Standardized API for patient and population services, Office of the National Coordinator for Health Information Technology. FHIR Release 4.0.1 at 170.215(a)(1), US Core STU 6.1.0 with STU 7.0.0 and 8.0.1 alternatives, SMART App Launch Release 2.0.0 at 170.215(c)(2), OpenID Connect Core 1.0, the three launch modes, and the March 2024 authorization revocation date.
- FHIR v5.0.0, HL7 International. The current published version label, the Release 5 Standard for Trial-Use designation, and HL7’s own one-line definition of FHIR.
- Information Blocking, Office of the National Coordinator for Health Information Technology. The definition, the named actors, the regulatory citation at 45 CFR 171.103, and the exceptions across 45 CFR Part 171.
- United States Core Data for Interoperability, ASTP/ONC. What USCDI is, the published versions through V6 in July 2025 with V7 in draft, and the Standards Version Advancement Process that moves the target.
M. Sabir Junaid
Founder • CEO
M. Sabir Junaid is founder and CEO of Velcod, the senior product studio. He leads product and delivery across MVPs, SaaS platforms, and mobile apps — system architecture, no-code and low-code engineering, and AI integration. An official Bubble partner, Sabir has shipped products across fintech, marketplaces, edtech, and creator platforms. He works with founders raising pre-seed to Series A who need a product built to launch and scale in weeks, not a six-month agency build.
Share
- Previous
- Next
Ready to build?
Your idea, live in 3 weeks.
Book a free strategy call and get a fixed-price roadmap for your MVP, SaaS, or app.