Embedded Agents

The idea

You build the agent in Phrony. Your product calls it. Your customers, your teams, or your systems talk to it — and every interaction comes back with the reasoning and the record intact.

One agent, every surface. The same agent powers your web app, your support chat, and your back-office tools.

Governed wherever it runs. Guardrails, policy checks, and records apply on every call — no matter who's calling.

Ready for your stack. REST today. SDKs for the languages your team already uses.

Ship faster with the official client

Call the same public API your dashboard uses: start runs, poll status, read the conversation, and stream events. Install @phrony/sdk from npm. Keep the API key on your server—never in a browser bundle.

your-backend.ts

import { Phrony } from "@phrony/sdk";

const phrony = new Phrony({ apiKey: process.env.PHRONY_API_KEY! });
const agentId = process.env.AGENT_ID!;
const { runId } = await phrony.startRun(agentId, {
  input: { query: "Hello from your product" },
});
const run = await phrony.getRun(runId);
// run.output when the run completes — same REST API, typed client.

Financial services

Embedded in your mobile app

A decision-capable agent inside your product.

Give your users an agent that doesn't just answer — it acts. It reads the context of what they're doing in your app, pulls the data it needs, executes the action, and returns a clear result. All inside the same governance layer your auditors already trust.

Example: A customer opens the app, taps "review my spending," and asks whether they can afford an upcoming bill. The agent reads their transactions, projects the month's outflows, confirms coverage, and offers to move money if needed — inside the app, in one interaction.

Insurance

Embedded in your underwriting platform

A risk assessor next to your underwriter.

When a submission arrives, the agent reviews it against your appetite, pulls the relevant history, and produces a structured recommendation with the factors it weighed. Your underwriter sees the reasoning, adjusts what needs adjusting, and moves on — reviewing far more volume without cutting corners.

Example: A mid-market property submission lands in the queue. The agent checks occupancy, loss history, and geography against your guidelines, flags two concerns, and proposes terms. The underwriter tightens a deductible and binds.

Healthcare

Embedded in your clinical operations tool

A specialist alongside your review team.

Hand the agent the repetitive reviews — prior authorizations, eligibility checks, documentation completeness. It works inside the system your team already uses, pre-approves the straightforward cases, and puts the complex ones on top of the queue with a recommendation and the reasoning attached.

Example: A prior-authorization team receives fifty requests a day. The agent reviews each one against payer policy and patient history, approves the clear cases, and surfaces three that need clinical judgment — with every factor visible.

Build an agent once. Embed it anywhere.

The idea

Ship faster with the official client

Put a reasoning agent where the decision actually gets made.

Embedded in your mobile app

Embedded in your underwriting platform

Embedded in your clinical operations tool