agent composer
Bring the agents you already built under governance — without rewriting them.
Waxell Agent Composer adapts your existing agent logic — and new intent — into deployable agents with policies, budgets, and telemetry wired in from the first execution. The source behavior is preserved. The output is governed, testable, and compatible with the Waxell runtime on its first run.
Free to start. 2-line setup.
With it and Without it
WITHOUT IT
You have agent code that works in a notebook. Getting it production-ready means manually wiring policies, retrofitting budget limits, and bolting on telemetry — weeks of governance plumbing per agent, bottlenecked on the one person who knows how the runtime works.
WITH IT
THE PROBLEM
Most teams have agent code they need to govern, not greenfield projects. And most build agents first and figure out governance later — policy assignments added after the fact, budget limits retrofitted, telemetry bolted on. That work concentrates on a handful of specialists, which slows iteration and makes every new use case a return trip to the same person.
What it Produces
Every Agent Composer run outputs a deployable agent wired to the Waxell runtime: code that references the correct tool calls and execution context; governance assignments that apply the relevant policies and budgets from the moment the agent runs; telemetry instrumented at each step; and a test-ready artifact that can be validated in Waxell Testing before it reaches production.
For migrated agents, the process adapts existing behavior against the runtime's governance and execution model. It works from a structured understanding of your tools, policies, and environment — it doesn't guess at how agents should behave, it composes them against the actual constraints of your system.
MIGRATION WITHOUT REWRITE
Existing experimental or production logic is adapted into the Waxell runtime — already compatible, no manual rewrite, no translation layer. Source intent preserved; output governed.
GOVERNED FROM THE FIRST RUN
Every agent inherits policies, budgets, telemetry, and execution constraints automatically. It cannot bypass the governance plane or introduce untracked behavior — so agent creation never becomes a parallel, ungoverned path.
ITERATE, DON'T CONVERT ONCE
Test runs execute against the same runtime used in production, with full telemetry and policy enforcement. Behavior can be inspected, adjusted, and rerun without touching production state.
POLICY A
POLICY B
POLICY C
POLICY D
Designed to scale
