Phrony<br>Agents as first-class primitives<br>Phrony is an open specification and open-source runtime for declaring, deploying, and running AI agents — the way teams already treat services and infrastructure.<br>View on GitHubRead documentation

A member of

agent.yamldemo/triage<br>model<br>1apiVersion: phrony.com/v1

2kind: Agent

3metadata:

4name: triage

5namespace: demo

6version: 1.0.0

7spec:

8purpose: Triage incoming claims…

9model:

10provider: anthropic

11name: claude-sonnet

12tools:

13- ref: claims-db.read-claim

14- ref: routing.assign-queue

15policies:

16- name: route-only-known-queues

17limits:

18max_tokens_per_run: 20000

19max_loop_iterations: 8

20hitl:

21- trigger: tool:routing.assign-queue

The paradigm<br>Agents outgrew the code they were born in<br>Every agent starts small: a prompt, a few tools, some glue inside an app. That's the right shape for a helper inside a feature. But the agents teams ship now have grown up — many tools, real budgets, escalation paths, humans in the loop, and constant change. The thing you care about is no longer a function. It's an entity. Code is the wrong place to keep it.

How it works<br>Declare it. Deploy it. Run it.<br>Phrony treats an agent as a first-class primitive — the way you already treat services and infrastructure.

01Declared→<br>02Deployed→<br>03Run

agent.yamlbank/payment-assistant<br>policies<br>1apiVersion: phrony.com/v1

2kind: Agent

3metadata:

4name: payment-assistant

5# …

12policies:

13- name: large-payment-approval

14scope: tool:submit_payment

15conditions:

16field:

17op:

18value:

19decision:

20type: require_approval

21# …

Declared<br>Its purpose, tools, policies, limits, and human checkpoints live in one versioned manifest, not scattered across application code.

Technically, this gives you one place where agent behavior is decided. For governance, it means every action passes through a single enforcement and evidence point — by construction, not by convention.

Built to be adopted, not locked in<br>Phrony is offered as a methodology and a runtime you can run yourself. The spec is the standard; the open-source project is the reference implementation.<br>Open specification<br>The manifest schema, policy model, runtime contract, and trace format are open. Anyone can implement a conformant runtime. Manifests are portable.

Open-source runtime<br>The reference implementation is on GitHub. Run it locally with Docker, validate manifests, deploy agents, and drive sessions with the operator CLI.

What the runtime handles<br>Session lifecycle, the model loop, tool dispatch, policy enforcement, limits, human-in-the-loop pauses, and structured traces — so you do not rebuild that stack in every service.

phrony-platform/runtimeDocumentation

A member of

© 2026 Phrony. Open specification and open-source runtime.