Engine registry
AGS ships as a catalog of engines. Each engine owns a semantic domain, registers MCP tools / gates, and participates in Sentinel reflexes when something fails. There are 24 engines in the public registry.
Start with Staging, Semantic, Pointer, Sentinel, Efficiency, and Versioning. Add Canonical + Data Arrays when docs and lists start drifting. Layer ops and trust engines as you promote to shared environments.
Core loop (wave A)
Classify, mean, guard, version, deploy, measure.
| Engine | Domain | What it does | Try |
|---|---|---|---|
| Staging | staging | Classify diffs into atomic Commit-Scopes before an agent commits. | Showcase |
| Semantic | semantic | Constraint graph over concepts — contradictions surface as animated conflicts. | Showcase |
| Pointer | pointer | Map repo paths to rules and detect documentation drift. | Showcase |
| Sentinel | sentinel | Pain signals fire reflex arcs that deny raw git, force MCP commits, and clear only after real remediation — consistency with fewer wasted retries. | Showcase |
| Versioning | versioning | Semantic version graphs, bump rules, and changelog animation. | Showcase |
| Deployment | deployment | Environment pipeline from dev → bench → staging → prod with health gates. | Showcase |
| Efficiency | efficiency | Token budgets, loop prevention, and organic guardrails that make agents cheaper over time. | Showcase |
| Impact | impact | Growth auditor that proves the agent is getting smarter — curve-aware maturity, adversarial stress, no-op detection. | Showcase |
Structure & policy (wave B)
Singularity, lists, deps, release, promotion.
| Engine | Domain | What it does | Try |
|---|---|---|---|
| Canonical | canonical | Highlight the canonical source and animate reference drift. | Showcase |
| Data Arrays | data-array | Normalize, order, and deduplicate governance arrays. | Showcase |
| Dependency | dependency | Dependency graphs with cycle detection and compatibility cues. | Showcase |
| Policy | policy | Browse .ags/*.yml policies and watch enforcement animate. | Showcase |
| Release | release | Inspect release artifacts, tags, and generated notes. | Showcase |
| Promotion | promotion | Gate promotions across environments with animated checks. | Showcase |
Operations (wave C)
Health, observability, rollback, artifacts, registry.
| Engine | Domain | What it does | Try |
|---|---|---|---|
| Health | health | Live health-check simulation with endpoint status pulses. | Showcase |
| Observability | observability | Logs + metrics viewer with anomaly detection animation. | Showcase |
| Rollback | rollback | Rollback flow with version tracking visualization. | Showcase |
| Optimization | optimization | Performance budgets and MCP cost visualization. | Showcase |
| Artifact | artifact | Artifact inspector with structure validation animation. | Showcase |
| Registry | registry | Private npm registry simulation and publish flow. | Showcase |
Trust boundary (wave D)
Integrity, identity, access, security.
| Engine | Domain | What it does | Try |
|---|---|---|---|
| Integrity | integrity | Checksum visualization and tamper detection. | Showcase |
| Identity | identity | Identity graph for agents, users, and environments. | Showcase |
| Access | access | Permission matrix with access-gating animation. | Showcase |
| Security | security | Secret scanning simulation and secure deploy cues. | Showcase |
How agents use them
Engines are not “pick one.” A governed commit typically touches several: staging classifies the
diff, confidence scores risk, domain engines validate touched surfaces, Sentinel aggregates halt
signals, then CAP allows commit_checkpoint. See architecture for the ordered pipeline and API for tool names.