01Brief
Bound the task
A useful run begins with a small, testable intent. The brief defines what the agent may change, what it must preserve, and when it should stop.
ProducesIntent brief
Dynamic workflows in a new dimension.
ReactorJet builds and shares practical ways for agent systems to observe what changed, check the result, and decide what should happen next.
learnGuides and field notes
22publishedreadings
Learn a workflow
Tutorials, practical guides, concepts, and field notes for people building and using agents.
buildComponent library
5documentedcomponents
Build with components.
Explore and build with a library of components for planning, policy, telemetry, security, and evaluation.
deployAutonomic workflows
3workingsystems
Intent to production
See loops, tools, and operating patterns turn a request into checked, deployable work.
The component library
Discover every step from intent to deployment.
See the requirements, controls, evidence, and handoffs behind each step, then follow how those responsibilities translate across tools and components.
Repository telemetry
Read the state before deciding what should move.
Candidate change
Keep every proposal bounded and reversible.
Held-out evaluation
Judge the consequence outside the proposer.
Policy boundary
Fail closed when evidence is incomplete.
Next-best action
Let the verdict shape the next iteration.
Evidence trail
Keep the decision and its proof attached.
Runtime signal
Surface the events that matter to the loop.
Trust envelope
Make autonomy narrow, explicit, and visible.
Regression budget
Measure what the candidate is allowed to spend.
Operator handoff
Bring people in at the consequential boundary.
Recovery path
Treat rollback as part of the workflow design.
Control surface
See every state transition as it happens.
Explore the control loop
Trace the decision, not just the output.
Follow a bounded agent task from its first brief to the final decision. Inspect each handoff, then switch the outcome to see where human judgment enters the loop.
Control plane
Work plane
Intent to deployment
Tell the agent what you want. Carry it to production.
A clear request becomes a bounded plan, implementation, independent checks, and a deployment decision. The control loop stays consistent even when the tools change.
controller.ts
import { closeLoop } from "@reactorjet/core";
await closeLoop({
observe: repositoryState,
propose: boundedCandidate,
verify: heldOutEvaluator,
adapt: nextBestAction,
});Autonomic workflow
Say what you want. Watch the loop do the rest.
The agent can choose the steps without losing the controls: scope stays bounded, evaluation remains independent, and consequential actions still wait for a clear decision.
Find out howBiro
A practical operating model for agent-driven work.
Biro is where this model is taking shape: one environment for expressing intent, directing agents, checking evidence, applying policy, and deciding what moves forward.
ReactorJet documents that work as it develops. As models improve and tooling matures, we are building and testing the controls, components, and workflows that make the model practical, then bringing them together in Biro.
New from ReactorJet
Ideas you can use in your next agent build.
Practical notes and honest teardowns about agent behavior, evaluation, safety, and control. New entries always end with something you can try.
- Builder note2026-08-13
Agent activity is not progress
Tool calls and busy run logs measure agent activity; progress begins when the workflow can prove that the target state moved closer and stayed there.
- agent-systems
- evals
- Builder note2026-08-13
Approval fatigue is a workflow design problem
When an agent asks for approval too often and without decision-ready context, the workflow trains people to click through instead of protecting the actions that deserve judgment.
- agent-systems
- trust-boundary
- Builder note2026-08-13
Rollback belongs in the plan
An agent should know how a consequential action will be contained or reversed before it acts, because a rollback invented after failure is only another untested production change.
- autonomic-swe
- agent-systems
FAQ
Questions, answered.
What ReactorJet is, what you can use now, and how to start building with the ideas.
What is ReactorJet?
ReactorJet is a public library for designing more reliable agent workflows. It brings together tutorials, reusable components, field notes, and working examples that connect intent, action, evaluation, policy, and deployment.
Can I use these components?
Yes. Use them as implementation patterns and adapt the interfaces to your stack. Each component explains the responsibility it owns, the evidence it expects, and the boundary it should not cross.
What makes a workflow dynamic?
It can observe the result of its own action, compare that result with an explicit target, and use the verdict to choose the next move. A fixed automation sequence cannot do that.
Does ReactorJet replace human review?
No. It makes the decision boundary explicit. People set goals, define trust boundaries, and decide which verdicts may advance automatically. The system keeps evidence attached to every consequential change.
Where do the evaluators run?
Outside the proposer whenever possible. Separating proposal from judgment reduces self-grading and makes failures easier to reproduce, audit, and contain.
Where should I start?
Start with one bounded task, one measurable target, and one independent check. Add policy, approval, and recovery only where the consequences require them.
From intent to deployment
Turn one clear request into checked, deployable work.
Learn how to bound the task, separate proposal from evaluation, and keep the evidence that decides what moves forward.