# ProDocuX AI compatibility

ProDocuX integrates with AI operating environments through explicit contracts rather than a particular model or vendor.

Interactive workbenches, autonomous runtimes, persistent workspaces, workflow platforms, and developer agents can keep their own interaction model. They connect through files, commands or skills, APIs, or tool protocols to a shared PDX support plane that provides orchestration, checkpoints, validation, diffs, evidence, manifests, and checksums.

## Responsibility split

- AI environment: interpretation, research, planning, and drafts.
- Structured contract: declared plan, inputs, assets, policies, expected formats, and approval rules.
- PDX Artifact Engine: controlled tool dispatch, checkpoints, resume, reconciliation, manifests, and receipts.
- ProDocuX Kernel: deterministic document intake, schema and rule validation, evidence capture, normalized diffs, rendering, and checksums.
- Human: professional judgment and final approval where required.

## Integration levels

1. Artifact handoff: export JSON, documents, or assets for PDX Core to process.
2. Command or tool call: invoke a deterministic command or skill.
3. Service integration: submit plans and receive artifacts plus verification results.
4. Orchestrated workflow: combine an optional planner, PDX Core, and human approval in a resumable flow.

## Boundary

PDX Core does not make LLM reasoning deterministic. For versioned inputs, rules, and skills, it makes execution reproducible and results independently inspectable. Invalid output becomes explicit machine-readable failure rather than a successful-looking artifact.

Human-readable page: https://prodocux.com/ai-compatibility