Healthcare AI Agents. Assembled.

From fragmented pilots to production-ready workflows, powered by open standards

Agents Assemble Challenge is on. Learn More →

Healthcare AI is everywhere.
Production is rare.

Everyone is building agents.
No one is assembling them into workflows

Fragmented agents. Isolated tools. Lost context.

Agents built by different teams can’t speak to each other.
Swapping a model means breaking and rebuilding the integration.
Patient context disappears across multi-agent workflows.
Compliance teams can’t trace what the AI did or why.

What is Prompt Opinion?

Prompt Opinion is an open standards platform for assembling context-aware healthcare AI agents into real clinical workflows

Healthcare AI doesn’t scale because intelligence and context live in different places. We connect them through open standards so that agents can be assembled, grounded, and deployed

Builders bring the intelligence.

Create agents and tools on open standards. Publish once. Make them available to every organization on the platform.

Organizations bring the context.

Connect your EHR, data, and policies. Discover agents. Assemble them into workflows grounded in your environment.

Open standards are the glue.

MCP, A2A, and HL7 FHIR ensure every agent remains interoperable, regardless of who built it or what model it runs on.

The platform makes it possible.

Authentication, credential bridging, context routing, audit trails, and governance handled once, for everyone. You focus on the problems worth solving.

What does Context-Aware Agent mean?

Every agent knows where it is working

Patient Context

At the bedside

One patient, one chart. The agent focuses on individual clinical decisions - drafting prior authorizations, summarizing visits, flagging risks for this person, right now.

Cohort Context

Across a population

Managing a panel or program. The agent shifts to group-level reasoning - finding care gaps, identifying patients with missing labs, surfacing trends across a population.

Research Context

In the evidence

Reviewing protocols, matching trials, and evaluating guidelines. The agent focuses on criteria and clinical evidence - verifiable, grounded, not invented.

Why is context important?

The agent knows how to behave based on where you are working and who is asking. Same platform, same agent, different context, but completely different output.

What You Can Assemble From?

Four sources. Any combination. One platform.

Every assembled agent draws from one or more of these four sources. Mix and match them to suit your use case, team, and context.

01

Platform Templates

Start Here

Pre-built agent templates configured for the most common healthcare workflows. Plug in your data, configure it to your context, and deploy. No build required.

02

Build Inside

Your idea. Your way.

Create a custom agent directly inside the platform. No external infrastructure needed. Define the use case, connect your data sources, and set the context. The platform handles the rest.

03

External Agents

Registry. Open Standards.

Builders publish agents to the registry on open standards. Discover them, connect them to your context, and assemble them into your workflows instantly, without custom integration.

04

MCP Tools & Servers

Plug & Play

Standalone tools published by builders include FHIR data layers, clinical calculators, coding lookups, and clinical trial databases. Any agent can plug into any tool. These are the building blocks every agent can use.

Assembly happens when you combine any of these sources with your organization's context.

One source or four. Internal only or fully external.

The platform connects them while keeping them grounded in your EHR, your policies, and your data.

What can Assembled Agents do?

See what assembled agents deliver today.

These are not demos. These are agents running on the platform right now, producing real deliverables. Each one shows what happens when context-aware AI is assembled with your data and connected to your workflows.

Each scenario below draws from a different combination of the four sources. These are not the only things you can build. They are examples of what becomes possible when builders and organizations come together on open standards

Prior Authorization

Context: Patient

Produces: Complete prior auth packet — evidence extracted from chart, payer form filled, submitted via API, follow-up task

For: Care teams, Revenue Cycle Management

Clinical Trials Matching

Context: Patient

Produces: Shortlist of eligible trials matched to a specific patient, extracted from unstructured notes against complex inclusion criteria.

For: Clinicians, Research Teams

Clinical Decision Support

Context: Patient / Cohort

Produces: Evidence-backed clinical brief grounded in your organization's guidelines, not a generic web search.

For: Clinicians, Clinical Leaders

Population Health Gaps

Context: Cohort

Produces: Table of patients with open care gaps sorted by risk, pulled from unstructured notes and structured EHR data.

For: Pop Health Managers, Ops Teams

Policies & Procedures

Context: General

Produces: Policy-aligned care guidance grounded in your organization's verifiable documents. Not invented.

For: Compliance, Clinical Leadership

Patient Summarization

Context: Patient

Produces: Longitudinal patient summary across years of chart history, with flagged anomalies and care continuity gaps.

For: Clinicians, Care Coordinators

Each of these agents is open for builders to extend, improve, or recreate through the

Agents Assemble Challenge

The platform is live. The ecosystem is being built right now.

Who is this for?

Builders and Organizations. Two sides of the same platform.

Prompt Opinion connects two groups who need each other. Builders who create and publish agents. Organizations that assemble and deploy them.

The platform is what makes both sides possible.

Builders


Create and publish to the ecosystem

A Builder is anyone with an idea to fix a healthcare problem. Clinicians, Informaticists, Ops leaders, and Engineers, all have a place here. You don’t need to write code to start. You need a use case worth solving.


What builders get

  • → Open standards infrastructure - the plumbing is done
  • → Reference implementations in TypeScript and Python
  • → Publish once, reach every organization on the platform
  • → MCP for tools, A2A for agents. You choose the layer
  • → Agents Assemble Challenge. $25,000 in prizes

Organizations


Assemble and deploy into your workflows

You know the problems. You have the data. You need AI that actually works inside your clinical workflows - not another pilot that never reaches production. Prompt Opinion gives you the infrastructure to assemble agents that are grounded in your context from day one.


What organizations get

  • → Agents connected to your EHR from day one
  • → Context-aware: patient, cohort, or research mode
  • → Discover from the registry or build your own
  • → From prototype to production without rebuilding infrastructure
  • → Audit trail, governance, and compliance built in

Prompt Opinion brings healthcare teams and agent builders onto the same platform, aligned around outcomes, governance, and real-world use.

Built on Open Standards

No lock-in. Any agent. Any model. Any source.

Prompt Opinion is built on three protocols that are becoming the global defaults for interoperable AI - MCP, A2A, and FHIR

Open standards mean any compliant agent or tool works, regardless of who built it, what model it runs on, or where it is deployed.

MCP

MCP

Model Context Protocol - Anthropic

The tools layer.
How agents discover and call external functions - FHIR queries, coding lookups, scheduling, document retrieval. Stateless and language-agnostic.
A2A

A2A

Agent-to-Agent Protocol - Google

The agents layer.
How autonomous agents communicate, delegate, and collaborate on complex multi-turn clinical tasks across organizational boundaries.
 HL7 FHIR

HL7 FHIR

Healthcare Data Standard

The healthcare data layer.
The industry standard for how patient data is structured and accessed across EHRs, labs, and payers. The foundation for clinical context.

The 5Ts: From Intelligence to Finished Work

AI shouldn’t stop at insight. It should end in action

Generative AI is powerful because it is non-deterministic. It can reason across ambiguity, adapt to context, and synthesize complex clinical information. But healthcare systems cannot run on probabilistic answers alone. If AI is going to move beyond pilots, it must be anchored to outcomes that are defined, auditable, and actionable.

That’s why we’re building around the 5Ts.

Not as a finished product, but as a principle. A way of designing agents so that flexible reasoning leads to a clear, intentional deliverable. A consultation. A document. A structured table. A completed transaction. A managed task.

The intelligence may be non-deterministic.
But the purpose must be clear.

TALK

The Consultant

Grounded reasoning across your data. A clinical consultation backed by your EHR, your policies, and your evidence; not a generic internet search.

TEMPLATE

The Scribe

Documents auto filled and ready to submit. Referral letters, prior auth packets, and discharge summaries - from blank page to done.

TABLE

The Extractor

Structured data pulled from unstructured sources. Scanned PDFs, clinical notes, and external records - all turned into clean, usable rows for analysis.

TRANSACTION

The Doer

Actions completed in connected systems. The agent stops being a viewer and starts submitting forms, creating orders, and scheduling appointments.

TASK

The Manager

Workflow management and follow-up. Nothing falls through the cracks. Persistent tasks, handoffs, and patient instructions are managed automatically.

The 5Ts define the deliverable. Next step is building the agents that deliver it.

Agents Assemble Challenge

The platform is live. The ecosystem is open. It's time to assemble.

Through the Agents Assemble challenge, we’re inviting builders to explore this idea with us, to design agents that combine generative reasoning with defined outputs. To test how determinism and non-determinism can work together in real healthcare workflows.

This isn’t a closed system.
It’s an open ecosystem.
And we’re building it together.

If you believe AI should deliver more than chat, build with us.

Design agents for real outcomes.
Publish them on open standards.
Help define what production-ready healthcare AI looks like.

Agents Assemble Challenge begins on March 4th

Talk to us

Tell us what you’re building.

We’ll help you get started quickly with Prompt Opinion.