BROWSER IDE
Run real agent logic in a browser — before it can touch anything.
Waxell Browser IDE is a browser-based environment for building and testing AI agents before they reach production. It runs real agent logic against controlled conditions — no installation, no infrastructure setup, no production connections — so an experiment can never quietly become a live system. When an agent is ready, promoting it into the Waxell runtime is an explicit step.
Free to start. 2-line setup.
THE PROBLEM
Most teams either prototype agents too casually or operationalize them too early. Lightweight demos produce brittle systems that fall apart under real conditions. Heavyweight setups slow iteration and discourage exploration. Teams put off serious experimentation because setup is expensive and mistakes are costly — so they learn how an agent really behaves only after it's already in production.
What It Is
A deliberately constrained space for running real agent logic — not a playground, not a production environment. Workflows behave consistently, but side effects and production dependencies are intentionally removed. Teams see how agents actually behave before those agents are allowed to touch live systems, with no installation and nothing to misconfigure.
OPEN IN A BROWSER
No installation, no infrastructure, no production connections. Sign up and start running real agent logic against controlled conditions immediately.
SERIOUS, NOT A SANDBOX
Run real agent logic and observe the full execution trace — which steps fired, what each decided, how the workflow resolved. Mistakes surface here, cheaply and visibly, before they reach systems that matter.
PROMOTION IS EXPLICIT
Agents get no access to production systems, persisted state, or operational controls by default. Moving into the Waxell runtime is always a deliberate step — no prototype silently becomes a production system.
The Browser IDE is the first stage of the Waxell lifecycle. Open it in any browser, load or define a workflow, run it against test inputs, and watch the full trace. Adjust parameters and rerun until the behavior is right. When the agent is ready, promote it into the runtime — at which point it inherits Waxell's governance plane (policies, budgets, telemetry, execution constraints) and is governed, testable, and accountable from its first real execution.
POLICY A
POLICY B
POLICY C
POLICY D
Designed to scale