Research Note

Specification-Driven Development Market Analysis

Enterprise Platform vs. Standalone IDE: AWS Kiro vs. GitHub + Microsoft Ecosystem

Paula Silva — Software GBB Americas | Microsoft Latam GBB
March 2026
Page 3 / 7

3. Approach Comparison: SDD Workflow Layer

At the SDD workflow layer specifically, both tools implement the same core pattern (specify → plan → decompose → implement) with different design philosophies.

AWS Kiro — Integrated IDE

Architecture: VS Code fork bundling SDD directly into the IDE. Three-file model: requirements.md (EARS notation) → design.md → tasks.md. Claude Sonnet via Bedrock. Agent hooks for event-driven automation. Steering rules. Multi-root workspace support. Background automation.

What It Does Better

  • Lowest friction to start — turnkey, zero configuration
  • EARS notation produces structured, testable requirements
  • Property-based testing (PBT) auto-generated from specs with shrinking
  • Checkpointing — rollback to any prior agent step
  • Agent hooks enable IDE-level file-save automation
  • Autonomous agent can work for days on complex tasks
  • Intuitive three-file model — easy mental model

Where It Falls Short

  • Single model provider (Claude via Bedrock) — no model choice
  • No security scanning, code review, or secret detection built-in
  • No governance, compliance, or content safety layer
  • No agent orchestration beyond the IDE
  • No enterprise data integration (M365, Teams, etc.)
  • No cross-agent memory or knowledge sharing
  • No agent SDK for custom agent development
  • Disproportionate overhead for small changes (Fowler analysis)
  • VS Code fork risk: may lag upstream extensions

GitHub Spec Kit + Platform — Composable Ecosystem

Architecture: MIT-licensed Python CLI (specify) that installs SDD slash commands into 25+ AI agents. Gated workflow: Constitution → Specify → Clarify (optional) → Plan → Analyze (optional) → Tasks → Implement. Constitution.md for immutable principles. Works with Copilot, Claude Code, Gemini CLI, Cursor, Windsurf, Qwen Code. Surrounded by full enterprise platform stack.

What It Does Better

  • Agent-agnostic: 25+ platforms, including Claude Code
  • Model freedom: Claude, GPT, Gemini, Llama, DeepSeek via GitHub Models + Foundry
  • BYOK for cost control and regulatory compliance
  • Constitution concept provides architectural governance
  • Explicit gated phases with human checkpoints
  • Integrated into GHAS security, Defender, Purview governance
  • GitHub Agentic Workflows for CI/CD-native automation (50+ templates)
  • Copilot coding agent for autonomous PR creation
  • Work IQ for M365 enterprise data access
  • Agentic memory (7% PR merge improvement, statistically significant)
  • MIT license — zero SDD-layer cost
  • Multi-variant exploration from single spec

Where It Falls Short

  • Verbose markdown output creates review burden (2,577 lines in one test)
  • No dedicated admin console for SDD governance
  • Execution quality varies by underlying agent
  • No built-in agent hooks (relies on GitHub Actions)
  • Higher learning curve (CLI + constitution + phases + platform services)
  • Spec Kit itself is experimental, seeking community feedback

Other SDD Tools & Frameworks

Beyond Kiro and Spec Kit, a broader ecosystem of SDD approaches is emerging. Each addresses different organizational needs and maturity levels.

BMAD-METHOD
Multi-Agent SDLC

Multi-agent SDLC orchestration with named personas (Business Analyst, Architect, Developer) and file-based context passing. Most comprehensive scope. Best for large greenfield projects. Highest adoption overhead.

Tessl Framework
Spec-Anchored

Spec-anchored/spec-as-source approach with explicit // GENERATED FROM SPEC code markers. Most advanced on Fowler's maturity scale. Maintains spec-code traceability through feature evolution. Early stage.

SPARC
Five-Phase Methodology

Five-phase methodology: Specification → Pseudocode → Architecture → Refinement → Completion. CLI v0.87.7 with multi-LLM support. Human-in-the-loop controls. Academic rigor. Niche adoption.

Ruflo
Multi-Agent Framework

Multi-agent orchestration framework for coordinating multiple AI agents on complex development tasks. Framework-level approach rather than methodology. Enables sophisticated agent collaboration patterns.

OpenSpec
Change-Centric Design

Change-centric, brownfield-first design. Proposal-first workflow. Strong fit for change-controlled industries with mandatory change documentation and audit requirements.

Ecosystem Note: None of these tools offer the platform integration, enterprise security, or scale of either Kiro or GitHub + Microsoft. Tessl's "spec-anchored" approach is worth monitoring — it represents a more advanced position on Fowler's maturity model than either mainstream option.

← Previous: Platform Architecture
Back to Report Index
Next: Capability Matrix →
Paula Silva — Software GBB Americas | Microsoft Latam GBB | March 2026