Schematic
drift: moderateA database management UI for Spring Boot (Java) that retrofitted the AI Unified Process artifact set — vision.md, requirements.md, entity_model.md, and fourteen use-case specs — onto a codebase started in 2022.
AI Unified Process (AIUP) emerging
Requirements-centric SDD that adapts the Unified Process for AI: use cases owned by requirements engineers, with code, tests, and docs regenerated around them.
Reviewed 3 months ago · Jul 18, 2026
Website GitHub · AI-Unified-Process/marketplace Compare with other frameworks →
- Stars
- 145
- Contributors
- 7
- Open issues
- 6
- Age
- 8 months
- Last push
- today
- Latest release
- aiup-core-v2.28.0 Oct 13, 2026
View weekly commit data
| Week starting | Commits |
|---|---|
| Oct 11, 2026 | 7 |
| Oct 4, 2026 | 3 |
| Sep 27, 2026 | 28 |
| Sep 20, 2026 | 0 |
| Sep 13, 2026 | 1 |
| Sep 6, 2026 | 0 |
| Aug 30, 2026 | 0 |
| Aug 23, 2026 | 3 |
| Aug 16, 2026 | 16 |
| Aug 9, 2026 | 2 |
| Aug 2, 2026 | 7 |
| Jul 26, 2026 | 21 |
| Jul 19, 2026 | 19 |
| Jul 12, 2026 | 14 |
| Jul 5, 2026 | 1 |
| Jun 28, 2026 | 0 |
| Jun 21, 2026 | 18 |
| Jun 14, 2026 | 5 |
| Jun 7, 2026 | 0 |
| May 31, 2026 | 0 |
| May 24, 2026 | 0 |
| May 17, 2026 | 0 |
| May 10, 2026 | 2 |
| May 3, 2026 | 2 |
| Apr 26, 2026 | 2 |
| Apr 19, 2026 | 4 |
| Apr 12, 2026 | 0 |
| Apr 5, 2026 | 2 |
| Mar 29, 2026 | 3 |
| Mar 22, 2026 | 0 |
| Mar 15, 2026 | 1 |
| Mar 8, 2026 | 1 |
| Mar 1, 2026 | 1 |
| Feb 22, 2026 | 12 |
| Feb 15, 2026 | 8 |
| Feb 8, 2026 | 4 |
| Feb 1, 2026 | 0 |
| Jan 25, 2026 | 1 |
| Jan 18, 2026 | 0 |
| Jan 11, 2026 | 0 |
| Jan 4, 2026 | 2 |
| Dec 28, 2025 | 0 |
| Dec 21, 2025 | 4 |
| Dec 14, 2025 | 4 |
| Dec 7, 2025 | 23 |
| Nov 30, 2025 | 3 |
| Nov 23, 2025 | 0 |
| Nov 16, 2025 | 0 |
| Nov 9, 2025 | 0 |
| Nov 2, 2025 | 0 |
| Oct 26, 2025 | 0 |
| Oct 19, 2025 | 0 |
Metrics updated Oct 15, 2026 · collected automatically from the GitHub API
Core approach
Adapts the Unified Process's four phases (Inception, Elaboration, Construction, Transition) to AI-native development. Use cases and entity models — not code — are the durable artifacts, specifying system behavior and maintained by requirements engineers for the life of the system. AI agents generate and regenerate code, tests, and documentation from them, with comprehensive tests protecting behavior across regenerations. Delivered as a Claude Code plugin marketplace: a stack-agnostic aiup-core plus technology-specific implementation plugins.
Workflow
/requirements— create the requirements catalog (aiup-core)/entity-model— create the entity model with ER diagrams/use-case-diagram— generate a PlantUML system use case diagram/use-case-spec— write a detailed use case specification/reverse-engineer— brownfield entry point: recover AIUP artifacts from an existing codebase/implement— generate the implementation from a use case (stack plugin:aiup-vaadin-jooqoraiup-angular-jpa)- Stack-plugin test commands (
/browserless-test,/playwright-test,/spring-boot-test,/vitest-test) lock in behavior before regeneration
Supported tools
- Claude Code
- Codex
- Cursor
- GitHub Copilot
- Gemini CLI
- OpenCode
Strengths
- Only framework tracked here rooted in the requirements-engineering tradition — specs describe system behavior and are owned by requirements engineers, not developer prompts discarded after implementation
- First-class brownfield workflow:
/reverse-engineerrecovers entity models and use case specs from existing code before iteration begins - Real enterprise adoption signals — named practitioner testimonials, a third-party-contributed stack plugin (
aiup-angular-jpa), and a companion IntelliJ plugin linking@UseCase-annotated tests to their UC-XXX specs - Distributed via open standards: Claude Code plugin marketplace plus
SKILL.mdskills installable through Tessl into Codex, Cursor, Copilot, Gemini CLI, and OpenCode
Limitations
- Young and small — the marketplace repo dates from February 2026 and has under 100 stars
- The full generate-and-test workflow is deepest on Java stacks (Vaadin + jOOQ, Spring Boot + Angular); other stacks get only the stack-agnostic core commands
- Unified Process ceremony (vision documents, BPMN, software architecture document, four phases) can be heavy for small tools and solo projects
- In-the-open dogfooding is still thin — the org's task-manager example carries a requirements catalog and entity model but only a single use case spec in-tree
Tracked projects using AI Unified Process (AIUP)
Notes
AIUP is the outlier among the frameworks tracked here: it comes from the requirements-engineering tradition (Jacobson-style use cases, the Rational Unified Process’s phase model) rather than from developer tooling. Its explicit critique of other SDD tools is that their specs describe the code — an elaborate prompt discarded once the feature ships — whereas AIUP use cases describe system behavior and outlive any implementation, with tests as the consistency engine that makes regeneration safe. The thing to watch is whether behavior-first specs maintained by requirements engineers survive contact with long-running maintenance better than developer-owned specs do; the public evidence base (two small example apps in the org) is still thinner than the methodology’s enterprise framing.