Why not just use an LLM?
LLMs can reason and generate. They do not own SAP readiness, authorization, release promotion or system truth.
See authority architectureCAPELLA AFEP / Agentic Full-Stack Engineering Platform for SAP
Govern AI-assisted SAP engineering from intent to verified outcome.
CAPELLA keeps qualification, human authorization, execution and SAP truth separate, so AI can accelerate SAP engineering without inheriting authority.
CAPELLA AFEP — Agentic Full-Stack Engineering Platform for SAP. Constellation particles build the CAPELLA AFEP wordmark, then reveal the governed authority chain: SFC qualification, GOVERN human authorization, A4Layer execution, SAP and OBSERVATORY evidence.
Current productCAPELLA AFEP 0.7.4
Release stateRELEASED
ComponentsSFC 0.13.1 / A4Layer 1.0.0
Latest engineering mainRTC + CI/CD evidence
Mission statement
CAPELLA exists because generation is not the same thing as governed SAP engineering.
We created CAPELLA to let AI accelerate SAP engineering without letting AI inherit engineering authority. It connects agentic reasoning, deterministic qualification, explicit human authorization, bounded execution, release control and SAP evidence into one inspectable journey.
Why not just use SAP tooling?
SAP remains final system authority. CAPELLA adds the governed journey around intent, qualification, human approval, execution and evidence.
Follow a governed missionWhy add CI/CD?
Engineering change must move through deterministic checks, trusted script verification, release control and environment evidence before it can be treated as production-ready.
Inspect release engineeringWhy keep proof separate?
A generated artifact, a deployed artifact and a SAP-verified outcome are different claims. CAPELLA keeps those boundaries visible.
Challenge the proofWhat CAPELLA governs
AI can create the candidate. CAPELLA governs what happens next.
Enterprise SAP engineering has three questions: can AI create it, is it qualified and authorized, and did SAP actually realize it? CAPELLA coordinates those transitions without allowing one state to impersonate another.
Can AI create it?
Generation creates a candidate. It is not engineering readiness.
Is it qualified?
Qualification checks scope, structure, preflight and evidence before approval.
Who authorized it?
Human authorization stays explicit and bound to the exact executable configuration.
Did SAP prove it?
SAP truth, readback and runtime evidence decide what can be claimed.
One governed mission
Ground, qualify, authorize, execute, verify.
This is the first-read model. The detailed lifecycle, authority labels and evidence views remain available inside Mission Control.
- GroundCAPELLA / ERCC
Turn an engineering request into bounded context and a clear task contract.
- QualifySFC
Check whether the candidate is technically ready to proceed.
- AuthorizeGOVERN
Bind a human decision to the exact executable configuration.
- ExecuteA4Layer
Carry only the authorized intent toward SAP.
- VerifySAP / Observatory
Separate realized system truth from activity, deployment and blocked runtime states.
CI/CD and release engineering
Change moves through trust gates before it can become production truth.
CAPELLA now exposes the release-management architecture behind governed engineering delivery: GitHub PRs, repository CI, signed-script verification, qualification, Release Train Controller reconciliation, human approval, environment promotion and Observatory evidence.
Post-0.7.4 engineering evidence, not a new product release.
Development 8795, Quality 8793 and Production 8791 remain distinct proof states.
Agents do not approve their own promotion or bypass signed-script trust.
What CAPELLA gives you
Six ways to see and control one governed mission.
Each surface leads with the user outcome first. The component and authority label stay visible as proof context.
Mission Control
Current state, approvals, operations, delivery and evidence timeline kept separate.
Open detailContext & Engineering Contract
Current Context, assumptions, unknowns, evidence and a frozen Engineering Task Contract made explicit.
Open detailQualification
SFC compile/qualification, typed field fidelity, SAP preflight and canonical artifact identity.
Open detailAuthorization
Separate SAP-write and Fiori-deployment authorization records.
Open detailGoverned Execution
Backend realization, service publication, OData verification, Fiori downstream deployment, reconciliation and runtime classification.
Open detailEvidence & Intelligence
Backend truth, operator truth, delivery verification, current/historical evidence, runtime evidence and Ask CAPELLA.
Open detailProof, not promises
Two evidence patterns, both bounded.
CAPELLA gains credibility by reporting success and blocked states with the same discipline. Historical proof stays historical; current-generation runtime blockers stay blocked.
Managed RAP Creation
A validated managed RAP mission moved through SFC 0.13.1 qualification, explicit human authorization, A4Layer 1.0.0 execution, SAP realization, SAP readback and ATC with 0 findings.
- Scope
- 14-object bounded SAP realization
- Boundary
- Historical bounded evidence for the supported managed RAP create path. It does not claim unrestricted RAP delivery, a clean test environment, or a new SAP realization in CAPELLA AFEP 0.7.4.
- SFC 0.13.1 qualification completed before authorization
- Explicit human authorization remained hash-bound
- A4Layer 1.0.0 carried bounded execution
- ATC 0 findings recorded in the historical evidence.
Governed FullStack deployment with platform-limited runtime
RAP-04 records a current-generation FullStack case where backend delivery and deployed content were verified, while runtime execution remained platform-blocked and was not represented as runtime-verified.
BACKEND: VERIFIEDFIORI_DEPLOYMENT: VERIFIED_DEPLOYEDCONTENT: 8/8 VERIFIEDSERVICE_ASSOCIATION: VERIFIEDAPPINDEX: VERIFIEDRUNTIME: PLATFORM_BLOCKEDRUNTIME_VERIFIED: FALSEMISSION_CLOSURE: PASS_WITH_PLATFORM_LIMITATION
Runtime execution proof is not established. SAP registration or deployment verification is not treated as runtime verification.
Inspect current-generation proofAuthority contract
Generate ≠ Qualify ≠ Authorize ≠ Execute ≠ SAP Truth
The public rule is simple. AI output, engineering qualification, human approval, execution and SAP truth are different states.
- Generation is not qualification
Generated managed RAP content is a candidate, not a qualified result. SFC 0.13.1 compiles and qualifies it before it can move toward a SAP-write authorization.
Choose your path
Understand CAPELLA, then challenge the evidence.
Start with the mission story or go straight to the proof ledger. Both paths keep authority and limitations visible.
Current status: E2E READY WITH BOUNDARIES. This is post-release engineering hardening and release-management evidence. It is not a new public product release, not a SAP realization claim, and not proof that production promotion or a fresh live mission completed.