Hi. Max here, one of the founders of Phrony. Emad built the runtime. We launched here before as a platform. Multi-agent orchestration, escalation, audit trails, the whole surface. That was the wrong thing to lead with. What people kept asking for was the layer underneath: something that actually runs the agent and holds it to a contract. That is what this launch is. Here is the thing we kept running into. Build an agent with a framework and the agent lives inside your application. That means your guardrails are prompts, your budget ceiling is a variable someone can edit, your approval gate is an if-statement, and your audit trail is log lines you have to reconstruct after something goes wrong. It holds together in a demo. It stops holding the moment the agent has real tools, real money and a human who has to sign off. "Please do not do X" is not a control. Neither is a check that lives in the same codebase as the thing it is supposed to check. Phrony moves the agent out of the application. You declare the tools, budgets, approval gates and policies in one manifest. The runtime executes that manifest and enforces it at dispatch, before the tool call happens, so there is no path around it. Every run comes back as a structured trace rather than logs. Infrastructure went through this exact transition. It lived in scripts and console clicks, then it became a declared artifact with its own execution engine, and nobody went back. Agents are at the same point. Easiest way to judge it is to run it. Self-host, five minutes, agent running with a trace at the end of it. Quickstart below. What we want from today: tell us where the manifest model breaks for your setup. If your agent does something a manifest cannot express, that is the most useful thing you can give us. Quick start: https://phrony.com/docs/quick-start

Phrony

Declare an agent in YAML. Run it outside your app code.

Declare an agent in YAML. Run it outside your app code.

Phrony

Previous Phrony Launches

Previous Phrony Launches