Unclassified

IntegrateWise / SpineWorkspace — Product & Platform Roadmap

Status: Maintained
Current framing: 2026-08-25
Owner: IntegrateWise
Product: SpineWorkspace

This roadmap tracks the implementation path from the current repository state toward the current product framing. It is a delivery map, not a replacement for frozen technical contracts.

North-star outcome

SpineWorkspace should let a person work across the software their business already uses from one contextual work environment while their Twin and bounded agents perform governed operational work alongside them. Spine preserves business context and continuity, the Bridge enables communication/execution, source systems remain operational endpoints, and completed work is re-ingested, reconciled and converted into better organizational knowledge.

End-state loop

Business Need / Goal
        ↓
Fragmented Business Work
        ↓
Connected Tools
        ↓
Preprocessor / Ingestion
        ↓
Spine Fabric
        ↓
Projection
        ↓
Human Workbench + Twin Workbench
        ↓
Shared Work / Allocation
        ↓
Governance
        ↓
Bridge: MCP + ACP + ADK + Orchestrator
        ↓
Execution
        ↓
Source Tools
        ↓
Re-ingestion
        ↓
Verification / Reconciliation
        ↓
Organizational Memory / Knowledge
        ↓
Next Work

Roadmap status model

  • DONE — implemented and evidenced in the repository.
  • ACTIVE — current implementation priority.
  • NEXT — ready after the active phase's exit criteria.
  • LATER — dependent on earlier phases or product validation.
  • HOLD — deliberately deferred; do not build opportunistically.

Phase 0 — Product framing and terminology

Status: DONE

  • Separate IntegrateWise (company/platform builder) from SpineWorkspace (product).
  • Lock integratewise.ai as the company/platform domain.
  • Lock spineworkspace.com as the product domain.
  • Establish Business Context as the top-level product concept.
  • Establish Spine Fabric as contextual business memory + operational continuity substrate.
  • Establish Human Workbench and Twin Workbench as distinct working surfaces.
  • Establish the Bridge as the communication/execution boundary around Spine.
  • Establish source → Spine → projection → work → execution → source → re-ingestion as the continuity loop.
  • Add the Product Framing Lock.

Exit: terminology and product/system narrative are stable enough to govern implementation and documentation.

Phase 1 — Business Context ingestion and synthesis

Status: ACTIVE

  • Inventory source systems and connector projections against actual code.
  • Establish the ingestion/preprocessor boundary.
  • Normalize provider data into canonical entities/relationships.
  • Preserve provenance, lineage and temporal validity.
  • Verify Triage classification and routing behavior.
  • Ensure Triage cannot directly promote canonical memory.
  • Establish governed promotion into Spine.
  • Make replay/idempotency and reconciliation evidence explicit.

Exit: fragmented business data can enter the system, be synthesized into connected context, and become governed Spine state without bypassing authority boundaries.

Phase 2 — Spine Fabric and schema authority

Status: NEXT

  • Align the Spine schema with the current business-context ontology.
  • Separate canonical Spine records from operational/sandbox working state.
  • Define the authoritative write paths and writer contracts.
  • Define loop-state representation and lifecycle transitions.
  • Define context assembly inputs/outputs.
  • Define memory, evidence, provenance and lineage relationships.
  • Ensure AI/runtime paths cannot write directly to canonical truth.

Exit: Spine is a coherent, governed continuity substrate with explicit canonical and operational boundaries.

Phase 3 — Projection engine and Human Workbench

Status: NEXT

  • Project the same organizational reality by role, domain, responsibility, authority and relevance.
  • Preserve cross-domain/cross-team context without duplicating truth.
  • Implement the Human Workbench information architecture.
  • Establish Today / Personal Work.
  • Establish Communications.
  • Establish Organization Memory / Knowledge.
  • Establish Work surfaces.
  • Add business-domain KPI and operational views through projections.
  • Ensure connected-system work can be performed from the contextual work surface without recreating every SaaS UI.

Exit: each user sees the right business context and work for their responsibilities while the organization retains one underlying truth model.

Phase 4 — Twin Workbench and Coworker

Status: NEXT

  • Establish one Twin per user/work identity.
  • Bind Twin context to the user's role/domain/authority projection.
  • Provide governed operational sandbox context to the Twin.
  • Implement Twin work planning, delegation, execution monitoring and result reporting.
  • Implement Coworker as a contextual companion to active Workbench interaction.
  • Ensure Coworker is not a second authority or orchestration engine.
  • Establish shared Work/Task assignment state between human, Twin and agents.

Exit: the user and Twin can work on the same business Work with clear authority, context and assignment boundaries.

Phase 5 — Bridge and agent communication

Status: NEXT

  • Define Spine communication contracts.
  • Implement governed human ↔ Twin ↔ agent context handoff.
  • Implement ACP semantics for agent/Twin handoffs.
  • Keep MCP scoped to tool/resource interaction.
  • Use ADK as the agent runtime/development mechanism.
  • Use the Orchestrator for durable coordination, retries, dependencies and bounded agent execution.
  • Preserve Work ID, authority scope, constraints and evidence requirements across handoffs.
  • Ensure all communication is grounded in governed context.

Exit: agents and platform services can collaborate over the Bridge without creating parallel truth stores or authority paths.

Phase 6 — Governed execution and source-system continuity

Status: NEXT

  • Resolve capabilities from actual connector/capability registry.
  • Apply Governance before consequential execution.
  • Execute through the appropriate capability/workflow/provider.
  • Capture receipts, evidence and audit records.
  • Reconcile external state against intended outcome.
  • Re-ingest resulting changes into Spine.
  • Surface loop state and failures to the Human Workbench/Twin.
  • Make retries, compensation and idempotency explicit.

Exit: execution is complete only when the expected business state is verified and the continuity loop is closed.

Phase 7 — Knowledge and organizational memory lifecycle

Status: NEXT

  • Distinguish current context, retained memory and continuity.
  • Convert validated outcomes into governed organizational knowledge.
  • Preserve decision/evidence lineage.
  • Make knowledge available through projections and future context assembly.
  • Keep learned patterns separate from authority and policy.
  • Ensure memory cannot silently become canonical truth.

Exit: processed work improves future organizational context without bypassing governance.

Phase 8 — Host-aware ecosystem, discovery and marketplaces

Status: LATER

  • Host-aware identity and connected integration discovery.
  • Reusable host connector projections without copying provider secrets.
  • Connector discovery and health.
  • Capability discovery.
  • Plugin marketplace.
  • MCP marketplace/pool discovery.
  • Activation and lifecycle management.
  • Versioning and compatibility rules.

Exit: a host can connect its existing software ecosystem once and SpineWorkspace can discover and govern the usable capabilities across that ecosystem.

Phase 9 — Platform economics and operational infrastructure

Status: LATER

  • Identity and tenant infrastructure.
  • Connectivity infrastructure.
  • AI/agent runtime infrastructure.
  • Execution/orchestration infrastructure.
  • Communication infrastructure.
  • Governance infrastructure.
  • Observability and operational controls.
  • Unit-of-operational-work metering and cost attribution.
  • Reliability/SLO model for the complete work loop.

Exit: IntegrateWise can operate the platform substrate reliably and attribute platform consumption to operational work.

Phase 10 — Validation, go-live and continuous adaptation

Status: LATER

  • End-to-end business-work-loop fixtures.
  • Projection correctness tests.
  • Governance/authority tests.
  • AI sandbox isolation tests.
  • Bridge protocol tests.
  • Source reconciliation tests.
  • Memory-promotion tests.
  • Human Workbench usability validation.
  • Twin Workbench execution validation.
  • Production readiness and operational runbooks.

Exit: a complete business Work can be demonstrated from need through outcome, reconciliation, memory and next Work.

Current priority order

1. Business Context ingestion / synthesis
2. Spine Fabric + schema authority
3. Projection engine
4. Human Workbench
5. Twin Workbench + Coworker
6. Bridge communication
7. Governed execution
8. Source reconciliation / re-ingestion
9. Organizational memory / knowledge
10. Host-aware ecosystem + marketplaces
11. Infrastructure economics / operations
12. End-to-end go-live validation

Non-goals for the current phase

Do not expand the product by:

  • creating a separate product for every business domain;
  • creating independent truth stores for individual users or agents;
  • giving AI direct unrestricted canonical-Spine authority;
  • turning Coworker into a second agent authority;
  • recreating every connected SaaS interface;
  • treating MCP, ADK or an agent framework as the product itself;
  • promoting unverified source/AI output directly into organizational memory;
  • adding new top-level architectural concepts when an existing contract can express the requirement.

Roadmap maintenance rule

Every material implementation change should update this roadmap when it changes:

  1. phase status;
  2. an exit criterion;
  3. a dependency;
  4. a frozen product boundary; or
  5. a production-readiness blocker.

Do not mark a phase complete because code exists. Mark it complete only when its exit criteria are evidenced by tests, runtime behavior, documentation and the relevant source-of-truth contracts.