Two runs, two paths.
The same request, run twice, takes two different paths. You chain non-deterministic models and tools, expect one method, and get a new one each time.
REUSABLE AGENT WORK
Agent work, minus the randomness and the cost.
Rote turns a successful agent run into a Play that is deterministic, cheaper in tokens, more secure, and easy to share across your team.
Does this sound familiar?
Every run pays for discovery again.
More tool calls and more turns than you planned, and a bill that changes from run to run.
A rule is not a guard.
You ask the agent not to write. You cannot guarantee it will not.
Good work stays where it happened.
Engineering keeps it. Sales, marketing, and ops start from zero.
Rote records the run, drops the failed paths, and keeps the successful one as a Play.
A Play is a saved, runnable unit of work. It follows the same method every time and costs a fraction of the tokens.
From execution to reusable method.
Do real work.
The agent reaches APIs, local processes, and browser interfaces inside a recorded workspace.
Keep evidence.
Each useful response becomes an immutable, queryable workspace artifact instead of disappearing into chat.
Compile what worked.
rote turns the successful trace into typed inputs, declared steps, dependencies, effects, and presentation.
Change the inputs.
Another permitted agent invokes the Play URI and follows the preserved method with fresh inputs.
A Play is inspectable before it is executable.
A Play declares parameters, tools, services, authentication, privileged requirements, effects, steps, dependencies, ownership, visibility, and version. It can render an honest human report, concise summary, or canonical JSON result.
Execution is declared.
The method does not depend on prose alone.
What worked is preserved.
A skill guides reasoning; a Play preserves a specific executable method.
The contract travels too.
Inputs, effects, evidence, presentation, ownership, visibility, and versions stay attached.
Your harness stays yours.
rote works across existing agents and harnesses and makes the validated method portable.
Repeat the work, not the reasoning.
Measured on modiqo/hello: the first frontier run spent 14,900 reasoning tokens. Every Play run after spent about 300. That is 98 percent less reasoning.
Predictable
Same steps, same tool path, every run.
Fewer tokens
About 2% of the reasoning, on a cheaper model.
Less risk
No room to wander. It cannot write what it did not declare.