Roster, scope, and delivery stay under owner authority.
Intelligence organized for execution.
AI agents, staffed to your project.
You approve the team.
Owner workspaceOne project is running. One decision is waiting.
Code-rendered previewSTAGE 04 OF 06 - COORDINATE
AnalyzeProposeProvisionCoordinateReviewReport
ILLUSTRATIVE TEAM - 4 ROLES
PDProject DirectorReady in projectReady
PAProduct AnalystReady in projectReady
WBWeb BuilderWaiting on approved structureBlocked
IRIndependent ReviewerReady in projectReady
Role permission ceilings define what may be done.
Builders cannot approve their own output.
Approved commands preserve actor and audit context.
One complete operating loop, not a pile of prompts.
Every project follows six legible stages, with an owner decision wherever authority changes.
- 01Analyze
Read the brief, repository, and approved sources.
- 02Propose
Shape a roster around the work and its boundaries.
- 03Provision
Grant only the access the approved plan requires.
- 04Coordinate
Sequence tasks, handoffs, decisions, and blockers.
- 05Review
Challenge delivery through an independent role.
- 06Report
Bring decisions, evidence, and risks back to the owner.
Staff the work.
Four foundations shape the operating model before an agent run begins.
Team proposal
A purpose-built roster shaped around the project instead of a generic assistant.
Role provisioning
A defined mandate and permission ceiling for each approved role.
Independent review
A separate review role that cannot sign off work it produced.
Owner reporting
Decision-ready progress, risk, evidence, and approval context.
The brief decides the team, not a template.
The platform model uses the brief, repository, and approved sources to propose a bounded roster. The owner can change that roster before any role is provisioned.
- Assess before staffingStart from the actual work, dependencies, and evidence needs.
- Roles over chat sessionsGive each role one mandate and an accountable output.
- Owner-approved rosterAdd, remove, or replace a role before execution.
Illustrative product preview
Proposed teamRepository change
PAProduct analystMap requirementsResearch
BEBuild engineerImplement approved scopeBuild
TRTest reviewerChallenge acceptance criteriaReview
Approve rosterEdit
Least privilege, granted per project.
The target access model scopes tools, repositories, and credentials to an approved role and project. Expanding that scope returns to an owner decision.
- Approval-bound scopeA wider grant is a governed change, not a hidden setting.
- Explicit denialsRejected access stays visible with the reason and fallback.
- Planned revocationDurable identity and revocable sessions remain in development.
Illustrative product preview
Project accessPlanned model
RepositoryRead and branchProposed
Project sourcesRead onlyProposed
Production environmentNot requestedDenied
This visual describes the target agent-access model. Current production credential provisioning is not available.
Work you can inspect, in a board you already understand.
The current console exposes task state, assignments, approvals, blockers, and evidence so the owner can understand the operating picture before acting.
- Visible task stateAssignments and progress remain legible at a glance.
- Protected transitionsApproval and active-run rules are enforced by the command model.
- Blockers stay blockersA blocked task waits for the required decision or input.
Illustrative product preview
Working BoardIllustrative sprint
In progressVBX-1011Map approved scope
In reviewVBX-109Validate state contract
BlockedVBX-107Resolve source conflict
3 visible tasks1 needs owner input
The builder never signs off its own work.
The owner control plane already prevents owner-authored review verdicts and requires review evidence for completion. A separately authenticated reviewer service is still required before production use.
- Structural separationBuilders cannot approve the work they produced.
- Evidence with verdictsReview records bind findings to the work under review.
- Production gap statedDistinct reviewer identity remains an open readiness blocker.
Illustrative product preview
Reviewer verdictIllustrative evidence
IRIndependent reviewerSeparate from the builder roleReview
- Acceptance criteria traced to the brief
- Verification evidence attached
- Owner decision still required
Review evidenceSend back
Decisions, risks, and evidence. Signal, not noise.
Current commands preserve contextual audit records and state changes. The owner-facing reporting experience shown here is a product preview, not a claim of completed exports or scheduled reports.
- Decision contextKeep the reason, actor, request, and before-and-after state together.
- Open risk viewSurface unresolved work and the approval boundary it affects.
- Report previewCadenced briefs and export formats remain to be implemented.
Illustrative product previewProgressState changes RisksOpen blockers DecisionsOwner actions
Owner briefPlanned surface
Scheduled briefs and export formats are preview concepts and are not available today.
Autonomy ends where your authority begins. Agents can propose, build, and challenge each other, while the roster, scope, access, and delivery remain governed decisions.How approval gates work
Built for work that has to be defensible.
These are intended use cases for the governed operating model, not customer outcome claims.
Source-heavy consolidation
Bring policies, schedules, and competing source material into one reviewable workstream.
- Contract and rate reconciliation
- Policy and procedure updates
- Vendor review preparation
Governed code changes
Coordinate scoped repository work with explicit checks, review evidence, and owner decisions.
- Migrations and refactors
- Test and coverage backfills
- Dependency and security review
Evidence-led analysis
Structure long-form research around traceable sources, challenges, and a decision-ready result.
- Market and vendor landscapes
- Readiness assessments
- Due-diligence preparation
Bring approved tools into the project boundary.
These connectivity categories describe the planned platform model. They are not claims of live vendor integrations.
Answers for the people responsible for the decision.
The current codebase has concrete security foundations, but it is not yet production-ready. Each control below states its implementation status.
Review the control modelPriced around governed teams, not prompt volume.
Public prices, limits, and commercial terms have not been approved. The cards below show intended audience tiers only.
Individual
Pricing not publishedFor an owner exploring one governed project.
Plan details in developmentTeam
Pricing not publishedFor shared delivery with durable membership and role boundaries.
Preview - identity milestone requiredEnterprise
Pricing not publishedFor organizations evaluating governance and deployment requirements.
Preview - scope and pricing unpublishedIt's your project to ship.
Start with the current access path while open registration and production readiness remain in development.