Skip to content

Roadmap

Where AIDLC Collaborative is today and where it is heading.

Current status

Phase Status Notes
Inception Working Inception Agent, requirements, user stories, tasks, Q&A, collaborative editing
Construction Working Construction Agent, ECS workers, code file tracking, real-time status, parallel dispatch
Review Working Blind and full review agents, modify agent, manual comments, pass/fail

What has been built

Inception

  • Free-form project descriptions as input (greenfield and brownfield)
  • Inception Agent that generates requirements, user stories, and tasks
  • Structured Q&A for clarifying ambiguities with mandatory ambiguity detection
  • Real-time collaborative editing of all artifacts (Yjs CRDT)
  • Graph-based traceability between requirements, stories, and tasks

Construction

  • Construction Orchestrator with parallel task dispatch
  • Multiple agents working simultaneously on independent tasks (up to maxParallelAgents)
  • Code file artifact tracking (path, commit, summary)
  • Follow-up questions during implementation
  • Real-time agent status streaming via WebSocket
  • Automatic PR creation on completion

Review

  • Blind Review Agent (evaluates without file details)
  • Full Review Agent (evaluates with complete context)
  • Risk scores and structured comments
  • Manual comments on both review types
  • Modify Agent for iterative refinement
  • Pass/fail decision workflow

Infrastructure

  • Multi-org, multi-project RBAC
  • AWS Cognito authentication
  • GitHub integration (OAuth, issue creation, status sync)
  • Neptune graph database for traceability
  • DynamoDB for operational state
  • Cross-sprint knowledge carry-forward
  • Terraform for full AWS deployment

What is next

Operate stage

  • Post-deployment monitoring integration
  • Error tracking and alerting
  • Feedback loop from production incidents back to specs
  • CI/CD pipeline integration

Agent improvements

  • Agent cost tracking and budget limits
  • Agent self-feedback after task completion
  • Review feedback extraction into reusable rules
  • Deeper brownfield support with semantic codebase indexing

Platform features

  • Notification system
  • Dashboard with project metrics and traceability reports
  • Audit log for all actions
  • Search across sprints and artifacts
  • Orphaned work detection (tasks without requirement justification)
  • Incomplete implementation detection (requirements without code)

Native integrations

  • Atlassian (Jira, Confluence) — sync requirements and tasks bidirectionally
  • GitLab — full Git integration parity with GitHub support
  • Other enterprise tools — based on user and customer demand

Alternative compute and data surfaces

  • Bedrock AgentCore as an alternative agent runtime
  • Vector databases for semantic codebase search (complementing the graph for backward traceability: code → purpose)
  • Alternative graph engines — the platform is not locked to Neptune; any graph DB that models the AI-DLC hierarchy works

Tooling evolution

  • Continued integration with tools adopted by users: Claude Code, OpenCode, Kiro, and emerging AI coding tools
  • The platform remains tool-agnostic — we integrate what works best for each module