Business knowledge you keep.
Capture the lifecycle in reviewable YAML or JSON. The next developer, operator or agent can inspect the same definition.
Make your workflows explicit.
Improve them with AI.
Define states, decisions and actions in versioned configuration. Give people and agents a shared model of how work progresses—and a clear starting point for the next improvement.
Independent Python package available · Connected Optophi workflow authoring is in development
SAME REQUEST. REVIEWABLE CHANGE.Improve the rule.
Improve the rule.
Understand the consequence.
A purchase request moves from draft to submitted, then to review or approval. Conditions check the information and approval requirements at each transition. Actions describe the work to perform along the way.
An agent can help propose a new approval threshold. Your team can inspect the definition change and evaluate the same cases under both versions before choosing whether to adopt it.
Approval is a lifecycle decision. Ordering and confirmed delivery are separate steps requiring the application's trusted integrations.
The same request$750 · equipment purchase
Required details are complete. Budget is available.
V1 · ORIGINAL RULE
Automatic approval up to $500
Awaiting reviewV2 · PROPOSED RULE
Automatic approval up to $1,000
Eligible for approvalOne changed condition. A different next step. Compare scenarios before adopting a revision. No workflow or purchase runs on this page.
A SHARED MODEL OF THE WORK
Business logic your team can build on.
Start with one repeated process. Keep its meaning accessible as the people, agents and systems around it change.
Changes you can examine.
Version the definition, inspect the diff and evaluate representative transitions before adopting a new rule.
Progress with explicit conditions.
Describe states, events, guards and actions. Configure checks for the facts required before a case can move forward.
Continuity across handoffs.
With persistence configured, keep a case’s state across people, agent sessions and systems. A refund or onboarding case has a lifecycle of its own.
Operations you can inspect.
Use the optional runtime’s transition history and metrics to investigate progress, failures and time spent in each state.
A starting point in existing work.
Discovery can infer candidate transitions from observed state sequences. Review and compare them; observed behavior still needs a domain expert’s judgment.
TWO WAYS TO WORK WITH AGENTSEstablish the logic.
Establish the logic.
Participate in the work.
Agent reasoning and explicit lifecycle rules can work together. Your team owns the process definition and how it is applied.
Build and improve the workflow.
Ask an agent to draft a machine, explain a transition, propose an exception path or prepare scenarios. Review the changes as part of your project.
Describe → propose → review → evaluateContribute within the workflow.
An agent can gather information, investigate a case or propose a decision. Route its proposal through the configured runtime to evaluate the next transition.
Gather facts → propose a step → check conditionsChecks apply on the runtime path you configure. Actor identity, approval evidence and external actions need trusted application integrations. An agent-supplied approval flag alone is not authority.
START SMALL. KEEP THE DEFINITION.
From one transition to an operated lifecycle.
The Python engine computes a transition from the current state, event and context. Add the surrounding layers when your application needs them.
- 01 · DEFINE & EVALUATE
Start in Python.
Load a YAML or JSON machine and inspect transition results locally. The core computes the plan; your application owns persistence and action execution.
Follow the quickstart - 02 · INTEGRATE
Keep the case state.
Add the runtime and a state store for application-owned lifecycles. Bind the guards and actions your process needs.
Explore state stores - 03 · OPERATE
Add operational layers.
Use the optional API, UI and worker queue for shared access and queued work, with behavior qualified for your deployment.
Read the worker guide
Definition versioning is a foundation for deliberate change.
Automatic migration of active cases, safe rollback and complete replay guarantees need separate runtime support and qualification. This page illustrates a design change; it does not demonstrate those capabilities or a connected Optophi run.
PART OF THE OPTOPHI APPROACHRequirements for data.
Requirements for data.
Structure for progress.
PyCharter describes what data must satisfy. PyStator describes how a business lifecycle can advance. Use each independently, or connect them through supported application interfaces.