ONC’s new LEAP award winners take on open-source agentic AI and lab interoperability

On October 6, ONC named the two 2026 winners of its Leading Edge Acceleration Projects (LEAP) in Health IT awards, a funding program the agency has run since 2018. The awards total $1.5 million. MedStar Health Research Institute will use its award to build and test open-source agentic AI for managing clinical follow-ups. Regenstrief Institute will use its award to help small, independent laboratories adopt standard codes such as LOINC. ONC’s principal deputy national coordinator called the moment “an inflection point in healthcare.”

ONC requires a live clinical pilot, two-platform integration, and a public risk assessment

In its June funding notice, ONC asks for much more. It requires MedStar to release all the software it develops under the award as open source, with full documentation. Other organizations must be able to set up the agent on their own.

The agent must support FHIR and work with third-party applications from unaffiliated developers.

MedStar also has to test the agent in the real world. It must run a pilot in a live clinical setting and connect the agent to at least two health IT platforms. Then it must publish a risk assessment that answers four questions:

• Is the tool ready for production?
• Does the AI-enabled workflow beat a non-AI approach?
• Is the tool usable?
• Are the underlying agentic AI standards mature enough for healthcare?

ONC capped the award at $1 million over two years.

MedStar will test the agent in shadow mode before a pilot with real patients

MedStar calls its tool the Sovereign Execution Bridge. ONC describes it as open-source middleware that coordinates complex clinical follow-ups.

The tool does three things. It manages follow-up processes that run for a long time. It stores data where the patient controls who gets access. It issues tamper-evident certificates so others can verify how the AI routed its work.

MedStar will work in three stages. First, it will build a national taxonomy of clinical follow-ups and configure the tool around it. Next, it will run a shadow-mode human factors phase that the project calls “Training Wheels,” and it will measure the results against three baseline approaches. Last, it will validate the tool with real patients in a clinical pilot with a human gate.

Lab interoperability award targets wrong LOINC codes and a problem ONC cannot measure

The second award goes after lab test codes. ONC says in the funding notice that some labs pick LOINC codes that do not match the tests they ran. Many others use codes they defined in-house, and smaller labs do so more often.

Local codes lose their meaning outside the lab that created them. The system on the receiving end has to map or interpret each result by hand. ONC says that raises the risk of delays, fragmented records, and misread results, all of which can hurt patient safety.

ONC also admits it cannot measure the problem. The agency says surveys have not worked, partly because few people respond and many respondents do not know the terminologies.

Regenstrief will recruit at least three small, independent labs. It will assess each lab’s readiness, check its data quality, fix its terminology problems, and validate the results. Regenstrief already offers LOINC mapping tools, and ONC requires it to report where existing tools did not work for these labs and why the labs still rely on local codes. ONC capped the award at $500,000.

ONC funded two of three planned areas and has not said what happens to the third

ONC had planned three awards worth up to $2 million, and it made two. The third would have paid up to $500,000 to add community feedback on real-world FHIR endpoint behavior to ONC’s Lantern monitoring tool.

ONC said it received applications for the agentic AI and lab areas. It did not say why the Lantern area went unfilled, whether it will seek applicants again, or what happens to the money it set aside. The funding notice leaves room for a second try. It stays open to applications for three years, and it says ONC may issue future awards under it.

ONC estimates the two funded projects will run from 2026 to 2028. The first look at MedStar’s progress should come sooner. The notice requires a workshop or webinar for stakeholders on the project’s status within 12 months of the award.