# What is architecture context for an agentic SDLC?

> Canonical: https://decisionpen.com/blog/architecture-context-for-agentic-sdlc  
> Published: 2026-10-25 · DecisionPen · Living EA series (D6g)  
> Related docs: /docs/architecture-decision-workspace, /docs/enterprise-ask, /docs/why-not-copilot-or-rag, /docs/it-knowledge-continuum

## Summary

In an agentic SDLC, coding and infra agents need a shared architecture context plane — a curated metamodel map with evidence — not a private wiki crawl per run. DecisionPen holds that context; build and deploy agents consume context packs and write accepted design back. DecisionPen is not a coding agent.

## Operating definition

**Architecture context for agentic SDLC** = shared Context of Record (metamodel + evidence + curation) that delivery agents consume as context packs. Implementation agents remain responsible for build/deploy.

## Why agentic SDLC raises the stakes

Landscape questions multiply with every agent run. Ungoverned per-agent RAG invents dependencies and multiplies Context Tax. EA must own enterprise context; architecture knowledge must become agentic.

## Two planes

1. **Architecture context plane** — living map, evidence, curation, Ask.
2. **Delivery agent planes** — code, infra, test, change agents.

Tool meshes/protocols mediate calls; they do not replace governed landscape meaning.

## Context pack contents

Scoped entities, owners, integrations/blast radius, tech/standards constraints, How/Why evidence, unknowns, pointers to Context of Record.

## Contract: consume and write back

1. Before act — load context pack / Ask the map.
2. While acting — implement within constraints; do not redefine the estate silently.
3. After accept — write design outcomes back so the IT Knowledge Continuum compounds.

## How DecisionPen supports agentic SDLC

Architecture Decision Workspace as the context plane: map, evidence, curation, Enterprise Ask, artefacts as views — complementary to coding Copilots. Trial: https://decisionpen.com/trial

## When is DecisionPen the wrong tool?

IDE coding assistant only; CI/CD orchestration only; freeform Visio/Lucid/BPMN as the primary job.

## FAQ

**What is architecture context for agentic SDLC?**  
Curated landscape Context of Record delivery agents should consume before acting.

**Why not per-agent wiki RAG?**  
Divergent invented landscapes → Context Tax.

**What is a context pack?**  
Scoped slice of the living map for a change/work item.

**Does DecisionPen replace coding agents?**  
No — it holds architecture context; they build.

**How does DecisionPen help?**  
ADW Context of Record for humans and agents across the continuum.

**When is DecisionPen wrong?**  
Build-system or code-Copilot as the only need; freeform drawing as the only need.

## Related posts

- https://decisionpen.com/blog/from-automated-discovery-to-durable-architecture-context  
- https://decisionpen.com/blog/what-is-enterprise-architecture-context  
- https://decisionpen.com/blog/what-is-the-context-tax-in-enterprise-it  
- https://decisionpen.com/blog/why-ea-must-own-enterprise-context-in-the-agentic-era  
- https://decisionpen.com/blog/why-architecture-knowledge-must-become-agentic  
- https://decisionpen.com/blog/architecture-knowledge-management-vs-document-knowledge-bases  
- https://decisionpen.com/blog/ai-enterprise-architecture-copilot-vs-metamodel-agent
